U.S. patent application number 12/081559 was filed with the patent office on 2008-09-04 for methods involving artificial intelligence.
This patent application is currently assigned to L'OREAL S.A.. Invention is credited to Fouad Badran, Philippe Bastien, Olivier De Lacharriere, Sylvie Thiria.
Application Number | 20080215610 12/081559 |
Document ID | / |
Family ID | 29420639 |
Filed Date | 2008-09-04 |
United States Patent
Application |
20080215610 |
Kind Code |
A1 |
De Lacharriere; Olivier ; et
al. |
September 4, 2008 |
Methods involving artificial intelligence
Abstract
One aspect of the present invention relates to methods of
generating a profile data set. In an exemplary embodiment, data is
accessed and the accessed data is processed using a dynamic cluster
method, mobile center method, and/or a k-means algorithm, each
using neighborhood data. Other aspects relate to methods of
generating a diagnosis, advice, and/or other information. Further
aspects relate to dynamic surveying and systems.
Inventors: |
De Lacharriere; Olivier;
(Paris, FR) ; Bastien; Philippe; (Charenton le
Pont, FR) ; Badran; Fouad; (Paris, FR) ;
Thiria; Sylvie; (Paris, FR) |
Correspondence
Address: |
FINNEGAN, HENDERSON, FARABOW, GARRETT & DUNNER;LLP
901 NEW YORK AVENUE, NW
WASHINGTON
DC
20001-4413
US
|
Assignee: |
L'OREAL S.A.
|
Family ID: |
29420639 |
Appl. No.: |
12/081559 |
Filed: |
April 17, 2008 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
10446926 |
May 29, 2003 |
7366707 |
|
|
12081559 |
|
|
|
|
Current U.S.
Class: |
1/1 ;
707/999.102; 707/E17.001 |
Current CPC
Class: |
G06Q 30/02 20130101;
G06N 3/02 20130101; G16H 10/20 20180101; G06N 5/02 20130101 |
Class at
Publication: |
707/102 ;
707/E17.001 |
International
Class: |
G06F 17/30 20060101
G06F017/30 |
Claims
1-177. (canceled)
178. A method of generating advice, the method comprising:
accessing data organized by an artificial intelligence engine, the
data being about a plurality of groups of characteristics, wherein
the data comprises at least one link between at least a first group
of the plurality of groups and a second group of the plurality of
groups; receiving information reflecting that a subject exhibits
the first group of characteristics; and processing the received
information and the accessed data, wherein the processing generates
advice related to the subject's predisposition to exhibit the
second group of characteristics.
179. The method of claim 178, wherein the artificial intelligence
engine comprises a neural network.
180. The method of claim 179, wherein the neural network comprises
a Kohonen map.
181. The method of claim 179, wherein the neural network comprises
a neural clustering algorithm.
182. The method of claim 181, wherein the neural clustering
algorithm is chosen from a dynamic cluster method, a k-means
clustering algorithm, and a hierarchical clustering algorithm.
183-185. (canceled)
186. The method of claim 178, wherein at least one of the first
group and the second group comprises only a single
characteristic.
187. The method of claim 178, wherein the received information
comprises qualitative information.
188. The method of claim 178, further comprising binary encoding
the received information.
189. The method of claim 188, wherein the binary encoding comprises
at least one of unconstrained binary encoding, additive binary
encoding, and disjunctive binary encoding.
190. The method of claim 188, wherein the accessed data comprises
binary encoded data, and wherein the processing comprises selecting
a portion of the binary encoded accessed data most closely
resembling the binary encoded received information.
191. (canceled)
192. (canceled)
193. The method of claim 178, wherein the accessed data comprises
qualitative data.
194. The method of claim 178, wherein the accessed data comprises
binary encoded data.
195. (canceled)
196. The method of claim 194, wherein the binary encoded data is
binary encoded using at least one of unconstrained binary encoding,
additive binary encoding, and disjunctive binary encoding.
197. (canceled)
198. The method of claim 194, wherein the accessed data is in the
form of a profile data set representative of the plurality of
groups.
199-201. (canceled)
202. The method of claim 178, wherein the advice is related to the
subject being likely to exhibit the second group of
characteristics.
203. The method of claim 178, wherein the advice is related to the
subject being unlikely to exhibit the second group of
characteristics.
204-208. (canceled)
209. The method of claim 178, wherein the advice comprises
information about at least one product.
210. (canceled)
211. The method of claim 209, further comprising offering at least
one product for sale to the subject.
212. The method of claim 209, wherein the processing comprises
determining that the subject has a predisposition to exhibit a
condition, and wherein the at least one product comprises a product
for treating the condition.
213. (canceled)
214. The method of claim 178, further comprising accessing a
plurality of queries, and presenting to the subject a subset of
queries from the accessed queries, wherein for at least some of the
queries presented, the method further comprises selecting a next
query as a function of the subject's answer to a previous
query.
215. (canceled)
216. The method of claim 214, wherein at least one of the queries
prompts the subject to indicate whether at least some aspects of an
initial profile resemble at least some aspects of the subject.
217. The method of claim 178, wherein the method further comprises
presenting at least one query to the subject and wherein the
receiving information comprises receiving at least one answer from
the subject.
218-220. (canceled)
221. The method of claim 178, wherein the artificial intelligence
engine comprises at least one of a neural network, constraint
program, fuzzy logic program, classification program, and logic
program.
222-255. (canceled)
256. A system, comprising: a data processor; and a storage medium
functionally coupled to the data processor, wherein the storage
medium contains instructions to be executed by the data processor
for performing the method of claim 178.
257-262. (canceled)
263. A computer program product, comprising a computer-readable
medium, wherein the computer-readable medium contains instructions
for executing the method of claim 178.
264. (canceled)
Description
[0001] This application claims benefit of priority to U.S.
provisional patent application No. 60/383,812, filed on May 30,
2002.
FIELD OF THE INVENTION
[0002] Aspects of the present invention relate to methods,
combinations, apparatus, systems, and articles of manufacture for
generating a profile data set and/or for generating a diagnosis,
advice, and/or other information. In another aspect, the invention
may include dynamic surveying. Certain exemplary embodiments may
involve data organized by an artificial intelligence engine.
BACKGROUND OF THE INVENTION
[0003] Traditional diagnostic methods are often inconvenient, labor
intensive, and expensive because they require in-depth analysis by
trained experts. In the realm of personal diagnosis, for example, a
subject suffering from an illness or other health condition
typically has to visit a medical professional to diagnose his/her
condition. This may require scheduling an appointment with the
professional, traveling to the professional's office, waiting to be
seen by the professional, finally being examined, and possibly
returning to the office for subsequent examinations. The
examination itself may require the subject to answer many questions
and perform a large battery of tests before a proper diagnosis is
obtained. Many times this process may require more effort and be
more expensive than the subject is willing to accept. Accordingly,
the subject may choose to not bother with the diagnosis. Thus, a
convenient means for the subject to obtain a preliminary diagnosis
from a remote location without answering an overwhelming number
questions would be beneficial.
[0004] In aspects outside of personal diagnosis, traditional
analysis typically involves collecting empirical data, distilling
the data, and drawing conclusions from the data, wherein the
conclusions may be used in the future for diagnostic purposes. In
order to make correct diagnoses from these conclusions, often a
large quantity of initial empirical data may be required. Large
amounts of data compound the problem of distilling and analyzing
the data. As a result, highly skilled individuals must spend
valuable time collecting, organizing, and interpreting the data to
glean useful information from it.
[0005] In another diagnosis example in the field of chemical
analysis, a researcher searching for unique properties of chemical
compounds (or chemical compounds having such properties) may
conduct numerous experiments and collect a massive quantity of
data. In order for the researcher to extract useful information
from the data, he/she has to perform the tedious task of organizing
and evaluating the experimental results. Therefore, it would be
beneficial to provide a means to organize data quickly and
accurately. Once data is properly organized, the researcher may
find the data useful, in at least the chemical analysis example, to
diagnose properties of additional samples.
[0006] Although the foregoing background discussion is directed
primary to diagnostics, it will become apparent in the following
description that many aspects of the present invention have
applicability in fields other than those involving a diagnosis.
Accordingly, the background discussion should be considered to be
exemplary of a few of many possible background issues that could be
addressed.
SUMMARY OF A FEW EXEMPLARY ASPECTS OF THE INVENTION
[0007] Methods, combinations, apparatus, systems, and articles of
manufacture consistent with features and principles of the present
invention may generate a profile data set; generate a diagnosis,
advice, and/or other information; and/or perform dynamic
surveying.
[0008] One exemplary aspect of the present invention may include a
method of generating a profile data set. The method may comprise
accessing data and processing the accessed data using at least one
of a dynamic cluster method and a k-means algorithm to generate
profiles for the profile data set. The accessed data may comprise
qualitative data. The dynamic clustering method and/or k-means
algorithm may use neighborhood data.
[0009] A second exemplary aspect of the present invention may
include a diagnostic method. The diagnostic method may comprise
accessing data organized by an artificial intelligence engine,
receiving information reflecting that a sample exhibits a first
group of characteristics, and processing the received information
and the accessed data. The accessed data may be about a plurality
of groups of characteristics and may comprise at least one link
between at least the first group of the plurality of groups and a
second group of the plurality of groups. The processing may
generate a diagnosis reflecting the sample's predisposition to
exhibit the second group of characteristics.
[0010] A third exemplary aspect of the present invention may
include a dynamic survey method. The method may comprise accessing
data organized by an artificial intelligence engine, accessing
queries, and presenting to a subject a subset of queries from the
accessed queries. The answers to at least some of the queries may
be used to process at least some of the accessed data. For at least
some of the queries presented, the method may further comprise
selecting a next query as a function of the subject's answer to a
previous query.
[0011] A fourth exemplary aspect of the present invention may
include a method of generating a profile data set. The method may
comprise accessing data about a plurality of groups of
characteristics, processing the accessed data to generate binary
encoded data representing modalities of the characteristics,
processing the binary encoded data to generate profiles for the
profile data set, and assigning at least some of at least one of
the plurality of groups, the accessed data, and the binary encoded
data to the profiles to generate the profile data set. Accessing,
processing, and/or assigning may use an artificial intelligence
engine.
[0012] A fifth exemplary aspect of the present invention may
include a diagnostic method comprising accessing a plurality of
queries, presenting to a subject a subset of queries from the
accessed queries, receiving information reflecting the subject's
answer to each presented query, accessing data about a plurality of
groups of characteristics exhibited by a plurality of individuals,
and processing the received information and the accessed data. The
data may comprise at least one link between at least a first group
of the plurality of groups and a second group of the plurality of
groups. At least one query answer of the subject may reflect that
the subject exhibits the first group of characteristics. For at
least some of the queries presented, the method may further
comprise selecting a next query as a function of the subject's
answer to a previous query. The processing may generate a diagnosis
reflecting the subject's predisposition to exhibit the second group
of characteristics.
[0013] A sixth exemplary aspect of the present invention may
include a method of generating advice. The method may comprise
accessing data organized by an artificial intelligence engine,
receiving information reflecting that a subject exhibits the first
group of characteristics, and processing the received information
and the accessed data. The data may be about a plurality of groups
of characteristics and may comprise at least one link between at
least a first group of the plurality of groups and a second group
of the plurality of groups. The processing may generate advice
related to the subject's predisposition to exhibit the second group
of characteristics.
[0014] A seventh exemplary aspect of the present invention may
include a method of generating information related to at least one
blood characteristic. The method may comprise accessing data
comprising blood characteristic data and hair characteristic data
for a plurality of respective individuals, receiving information
reflecting at least one hair characteristic of a subject, and
processing the received information and the accessed data. The
processing may generate information related to the subject's
predisposition to exhibit at least one blood characteristic.
[0015] A further aspect may relate to systems including structure
configured to perform one or more methods disclosed herein.
[0016] Additional aspects of the invention are set forth in the
description which follows and, in part, are obvious from the
description, or may be learned by practice of methods,
combinations, devices, systems, and articles of manufacturer
consistent with features of the present invention. It is understood
that both the foregoing description and the following detailed
description are exemplary and explanatory only and are not
restrictive of the invention as claimed.
BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The accompanying drawings, which are incorporated in and
constitute a part of this specification, illustrate several aspects
of the invention and, together with the description, serve to
explain exemplary principles of the invention. In the drawings,
[0018] FIG. 1 illustrates an exemplary flow chart for a diagnostic
method consistent with features and principles of the present
invention;
[0019] FIG. 2 illustrates an exemplary survey consistent with
features and principles of the present invention;
[0020] FIG. 3A illustrates a first exemplary list of queries and
query answers consistent with features and principles of the
present invention;
[0021] FIG. 3B illustrates a first exemplary diagnosis for a first
subject consistent with features and principles of the present
invention;
[0022] FIG. 3C illustrates a second exemplary diagnosis for the
first subject consistent with features and principles of the
present invention;
[0023] FIG. 3D illustrates a second exemplary list of queries and
query answers consistent with features and principles of the
present invention;
[0024] FIG. 3E illustrates a first exemplary diagnosis for a second
subject consistent with features and principles of the present
invention;
[0025] FIG. 3F illustrates a second exemplary diagnosis for the
second subject consistent with features and principles of the
present invention;
[0026] FIG. 4 illustrates an exemplary flow chart for a method of
generating a profile data set consistent with features and
principles of the present invention;
[0027] FIG. 5 illustrates an exemplary table of modalities using
additive binary coding consistent with features and principles of
the present invention;
[0028] FIG. 6 illustrates an exemplary table of modalities using
disjunctive binary coding consistent with features and principles
of the present invention;
[0029] FIG. 7A illustrates an exemplary table of modalities for a
group of characteristics consistent with features and principles of
the present invention;
[0030] FIG. 7B illustrates an exemplary binary encoded data
consistent with features and principles of the present
invention;
[0031] FIG. 7C illustrates exemplary binary encoded data of a
plurality of individuals consistent with features and principles of
the present invention;
[0032] FIG. 8 illustrates an exemplary grid of profiles consistent
with features and principles of the present invention;
[0033] FIG. 9 illustrates an exemplary arrangement of profiles
consistent with features and principles of the present
invention;
[0034] FIG. 10 illustrates an exemplary center median calculation
consistent with features and principles of the present
invention;
[0035] FIG. 11 illustrates an exemplary indicator function
consistent with features and principles of the present
invention;
[0036] FIG. 12 illustrates an exemplary Gaussian function
consistent with features and principles of the present
invention;
[0037] FIG. 13 illustrates an exemplary grid of profiles with
assigned binary encoded data consistent with features and
principles of the present invention;
[0038] FIG. 14 illustrates exemplary merged profiles consistent
with features and principles of the present invention;
[0039] FIG. 15 illustrates an exemplary flow chart for a method of
dynamic surveying; and
[0040] FIG. 16 illustrates an exemplary system implementing
features and principles of the present invention.
DESCRIPTION OF EXEMPLARY EMBODIMENTS
[0041] Reference is now made in detail to exemplary embodiments of
the invention, examples of which are illustrated in the
accompanying drawings. Wherever possible, the same reference
numbers may be used in the drawings to refer to the same or like
parts.
[0042] The exemplary embodiments described herein are primarily
related to diagnostic methods involving subject individuals and
certain individual characteristics relating to beauty, personal
habits, blood chemistry, etc. However, it should be understood that
the method, in a broader sense, could involve numerous other types
of individual characteristics, and, furthermore the invention could
be practiced to provide diagnoses of samples other than
individuals. In addition, it should be understood that certain
aspects of the invention have applicability in fields other than
those involving a diagnosis.
[0043] One embodiment of the invention may include a diagnostic
method. The method may be used to generate a diagnosis reflecting a
sample's predisposition to exhibit a group of characteristics. The
sample may be at least one of a subject individual, chemical,
molecule, item, location, idea, and/or any other subject of
interest to be diagnosed. Characteristics may include physical,
medical, physiological, biological, chemical, molecular, beauty,
situational, and/or any other characteristic. A group of
characteristics may include one or more characteristics. The
diagnosis of the sample's predisposition may include a likelihood
(and/or unlikelihood) that the sample currently exhibits, has
exhibited in the past, and/or will exhibit in the future, one or
more characteristics in the group.
[0044] Examples of characteristics that could be involved in a
diagnosis of a subject individual may include fashion preferences,
demographic, nutrition, cosmetic usage, medical history,
environment, beauty product usage, lifestyle, name, age, birth
date, height, weight, ethnicity, eating habit, blood (e.g., type,
chemistry, etc.), hair condition, vacation pattern, geographic
location of the individual's residence, location, work, work habit,
sleep habit, toiletries used, exercise habit, relaxation habit,
beauty care habit, smoking habit, drinking habit, sun exposure
habit, use of sunscreen, propensity to tan, number of sunburns and
serious sunburns, dietary restriction, dietary supplement or
vitamin used, diagnosed condition affecting the external body, such
as melanoma, facial feature, family history, such as physical
characteristics about relatives of the individual (e.g., premature
balding, graying, wrinkles, etc.), external body condition, color
preference, clothing style preference, travel habit, entertainment
preference, fitness information, adverse reaction to products,
adverse reaction to compounds, adverse reaction to elements (e.g.,
sun exposure), body chemistry, purchasing habit, shopping habit,
browsing habit, hobby, marital status, parental status, number of
children, country of residence, region of residence, birth country
and region, religious affiliation, political affiliation, whether
the individual is an urban dweller, suburban dweller, or rural area
dweller, size of urban area in which the subject lives, whether the
individual is retired, annual income, sexual preference, or any
other characteristic.
[0045] Blood characteristics may include any condition related to
the circulatory system of a living being (e.g., heart attack,
stroke, blood pressure, anemia, blood chemistry, blood type,
etc.).
[0046] According to features and principles consistent with the
invention, an embodiment of the invention may access data about a
plurality of groups of characteristics, as illustrated at step 102
in the flow chart of FIG. 1. Accessing data may include
receiving/obtaining data from a database, data structure, storage
medium, survey, and/or any other mechanism or combination of
mechanisms. The accessed data may be raw data, such as data entries
from a database, preprocessed data, such as encoded raw data, or
any other form of data. "Accessing" data may include at least one
of acquisition via a network, via verbal communication, via
electronic transmission, via telephone transmission, in hard-copy
form, or through any other mechanism enabling acquisition or
reception of data. In addition, "accessing" may occur either
directly or indirectly. For example, receipt may occur through a
third party acting on another party's behalf, as an agent of
another, or in concert with another. Regardless, all such indirect
and direct actions are intended to be covered by the term
"accessing" as used herein.
[0047] Accessed data, for example, may take one of many forms. It
may simply be a checked box, clicked button, submitted form, or
oral affirmation. Or it might be typed or handwritten textual data.
Accessing may occur through an on-line form, e-mail, facsimile,
telephone, interactive voice response system, or file transfer
protocol transmitted electronically over a network at a web site,
an Internet Protocol address, or a network account. Data may be
accessed from a subject for whom information is sought, or an
entity acting on the subject's behalf. Receipt may occur physically
such as in hard copy form, via mail delivery, or other courier
delivery. "Accessing" may involve receipt directly or indirectly
through one or more networks and/or storage mediums. Examples of
storage media may include magnetic storage devices such as floppy
disks and hard drives, optical storage devices, such as compact
discs and digital video discs, organic storage devices, electronic
storage devices, random access memory, virtual memory, permanent
memory, printed media, and/or any other medium for storing
information.
[0048] The term "network" may include a public network such as the
Internet or a telephony network, a private network, a virtual
private network, or any other mechanism for enabling communication
between two or more nodes or locations. The network may include one
or more of wired and wireless connections. Wireless communications
may include radio transmission via the airwaves, however, those of
ordinary skill in the art will appreciate that various other
communication techniques can be used to provide wireless
transmission including infrared line of sight, cellular, microwave,
satellite, blue-tooth packet radio and spread spectrum radio.
Wireless data may include, but is not limited to, paging, text
messaging, e-mail, Internet access and other specialized data
applications specifically excluding, or including voice
transmission.
[0049] In some instances consistent with the invention, a network
may include a courier network (e.g. postal service, United Parcel
Service, Federal Express, etc.). Other types of networks that are
to be considered within the scope of the invention include local
area networks, metropolitan area networks, wide area networks, ad
hoc networks, or any mechanism for facilitating communication
between two nodes or remote locations.
[0050] It should be noted that the terms "network" and "neural
network", as used herein, have distinct meanings. A "neural
network" may be any organization of "neurons" associated with one
or more algorithms, and/or may have any other configuration known
in the art of artificial intelligence. Accordingly, a "neural
network" may or may not be associated with a "network".
[0051] The data accessed in step 102 of FIG. 1 may be data that has
been organized as a result of processing using an artificial
intelligence engine. As described in more detail below, one example
of a data organization includes a profile data set resulting from a
neural network using one or more types of algorithms. However, it
should be understood that there are many other alternative ways in
which the data may be organized.
[0052] The accessed data may be based on information previously
collected in any known manner. For example, when the method
involves diagnosis of one or more individuals, FIG. 2 illustrates
an exemplary survey that may be used to collect characteristic
information for a plurality of individuals. Results of the survey
for a plurality of individuals may be optionally encoded, organized
using an artificial intelligence engine, and stored in a database
or other storage mechanism that may be accessed, as described
above. The survey may contain queries about many characteristics of
an individual and may be completed over the Internet. The
characteristics for a particular individual may include any
characteristic as described above. In this example, the survey may
contain queries about gender, smoking habit, occurrence of hair
loss, occurrence of white hair (i.e., senescent hair, such as
whitened or gray hair), occurrence of dandruff, concern about hair
loss, concern about white hair, concern about dandruff, concern
about baldness, coarseness of hair, color of hair, frequency of use
of fingernail hardener, frequency of use of fingernail multivitamin
treatment, frequency of use of nail polish, frequency of use of
hair treatments, etc.
[0053] For each individual completing the survey of FIG. 2, there
could be additional information collected about the individual.
Such additional information could be information that is not
normally known by an individual and/or information that is only
known after performing a particular test, examination, and/or
analysis. For example, for each of the individuals completing the
survey of FIG. 2, there could be a collection of additional
information relating to blood chemistry, anxiety test score,
depression test score, and/or any other information normally
requiring some form of test, examination, and/or analysis in order
to be revealed. The additional information could optionally be
encoded and stored in a database or other storage medium that may
be accessed, as described above.
[0054] Some of the characteristics may be grouped into a first
group of characteristics (e.g., the first group of characteristics
may include gender, concern about hair loss, concern about white
hair, concern about dandruff, smoking habit, concern about
baldness, coarseness of hair, and color of hair). The grouping may
be performed through an expert determination, algorithm, artificial
intelligence, or other mechanism. The grouping may be performed
prior to accessing data, wherein accessing receives data reflecting
previously grouped characteristics. Alternatively, the grouping may
take place, at least partially, after the data is accessed. The
grouping may be used to isolate characteristics that may predict
the exhibition of other characteristics by the individual. For
example, the grouping may be performed by an analyst wishing to
diagnose a subject individual based on the first group of
characteristic information from other individuals.
[0055] Other characteristics may be grouped into a second group of
characteristics (e.g., the second group of characteristics may
include occurrence of hair loss and frequency of use of a
fingernail multivitamin treatment) as previously described.
Additional groups of characteristics may be formed and any group of
characteristics may contain the same characteristic as another
group of characteristics.
[0056] As stated previously, the data about the characteristics may
be from many sources other than or in conjunction with the survey
response (e.g., the data may include blood characteristics from a
medical databank) and the accessed data may be partially processed
using some mechanism, such as an artificial intelligence engine.
Partial processing may include binary encoding as described
later.
[0057] The accessed data may include at least one link between at
least the first group of a plurality of groups and the second group
of the plurality of groups. The link may indicate exhibition of the
first group of characteristics by a sample implies likely or
unlikely exhibition of the second group of characteristics by the
sample. For example, a link may exist between a first group of
characteristics listed in FIG. 3A and a second group of
characteristics listed in FIGS. 3B and 3C, such that a subject
individual exhibiting the characteristics in FIG. 3A may be
diagnosed to exhibit the characteristics in FIGS. 3B and 3C.
Alternatively, the link may indicate exhibition of the first group
of characteristics by a sample implies unlikely exhibition of the
second group of characteristics by the sample. For example, a link
may exist between a first group of characteristics listed in FIG.
3A and a second group of characteristics listed in FIGS. 3B and 3C,
such that a subject individual exhibiting the characteristics in
FIG. 3A may be diagnosed to not exhibit all the characteristics in
FIGS. 3B and 3C. Further, the link may indicate that any one group
or combination of groups of characteristics implies likely or
unlikely exhibition of any other group or groups of
characteristics. Also, the link may be accessed via any of the
mechanisms described above, previously generated, generated at a
time data is accessed, and/or stored via any of the storage media
listed above. As described later herein, the link may be generated
using artificial intelligence.
[0058] An embodiment consistent with features and principles of the
invention may receive information reflecting a sample exhibiting
the first group of characteristics, as illustrated at step 104 in
the flow chart of FIG. 1. The information may be received in any
manner as described above in association with data accessing. For
example, FIG. 3A illustrates an exemplary list of answers supplied
by a subject individual in response to a plurality of queries
relating to the subject individual's exhibition of some
characteristics. The queries may be generated using a dynamic
surveying method described later. The query answers could be
received in electronic form or the query answers may be received in
paper form.
[0059] As illustrated at step 106 in the flow chart of FIG. 1, the
diagnostic method may further involve processing the received
information (from step 104) and/or the accessed data (from step
102). The processing may be performed in any known manner
compatible with the present invention. For example, the processing
may be performed according to one or more algorithms, and/or the
processing might involve use of an artificial intelligence
engine.
[0060] Processing in step 106 may generate a diagnosis reflecting a
sample's predisposition to exhibit the second group of
characteristics. For example, the diagnosis may be that an
individual who exhibits a first group of characteristics (e.g.,
male gender, not concerned about hair loss, concerned about white
hair, not concerned about dandruff, no smoking habit, not concerned
about baldness, fine hair, and red hair) is likely to have
previously exhibited in the past, to be exhibiting at the present
time, and/or will be exhibiting in the future a second group of
characteristics (e.g., hair loss and often use of a fingernail
multivitamin treatment). Such a diagnosis may be presented to the
individual, to a health care professional, beauty care professional
or any other individual or entity.
[0061] Based on the diagnosis, one exemplary method may provide a
prompt for a further diagnostic examination of a subject individual
by a practitioner. Examples of practitioners may include a health
care provider, beauty consultant, beauty care provider, dietician,
medical care provider, etc.
[0062] Alternatively (or additionally), the method may provide a
prompt for an individual to alter some habit, lifestyle, personal
care, etc. In another alternative (or additional) feature, the
method may include informing the subject about at least one product
for use by the subject. The product may be chosen from beauty
products, health products, and medical products. Further, the
product may be offered for sale to the subject. If the diagnosis
reflects a predisposition for the subject to exhibit a condition
(e.g., cosmetic condition, health condition, medical condition,
etc.), then the product may include a product for treating the
condition.
[0063] The term "product", as used herein, generically refers to
tangible merchandise, goods, services, and actions performed. The
term "beauty product" includes any product used for beauty, any
beauty care product, any cosmetic product, and any similar product.
A beauty product may be a product as defined above for affecting
one or more external body conditions, such as conditions of the
skin, hair and/or nails. Examples of tangible merchandise forms of
beauty products include cosmetic goods, such as treatment products,
personal cleansing products, and makeup products, in any form
(e.g., ointments, creams, gels, sprays, supplement, ingesta,
inhalants, lotions, cakes, liquids, and powders.)
[0064] Examples of service forms of beauty products include hair
styling, hair cutting, hair coloring, hair removal, skin treatment,
make-up application, and any other offering for aesthetic
enhancement. Examples of other actions performed include massages,
facial rubs, deep cleansings, applications of beauty product,
exercise, therapy, or any other action effecting the external body
condition whether performed by a professional, the subject, or an
acquaintance of the subject.
[0065] The following is exemplary and non-exhaustive listing of a
few beauty products: scrubs, rinses, washes, moisturizers, wrinkle
removers, exfoliates, toners, cleansers, conditioners, shampoos,
cuticle creams, oils, and anti-fungal substances, anti-aging
products, anti-wrinkle products, anti-freckle products, skin
conditioners, skin toners, skin coloring agents, tanners, bronzers,
skin lighteners, hair coloring, hair cleansing, hair styling,
elasticity enhancing products, agents, blushes, mascaras,
eyeliners, lip liners, lipsticks, lip glosses, eyebrow liners, eye
shadows, nail polishes, foundations, concealers, dental whitening
products, cellulite reduction products, hair straighteners and
curlers, and weight reduction products.
[0066] "Advice" includes one or more of product recommendations
(e.g., product recommendations for products to treat conditions),
remedial measures, preventative measures, predictions, prognoses,
price and availability information, application and use
information, suggestions for complementary products, lifestyle or
dietary recommendations, or any other information intended to aid a
subject in a course of future conduct, to aid a subject in
understanding past occurrences, to reflect information about some
future occurrences related to the subject or to aid a subject in
understanding one or more products, as defined above.
[0067] One example of the method illustrated in the flow chart of
FIG. 1 may involve accessing data from a database of individuals'
characteristics, wherein there may be links between groups in the
characteristics. Information about characteristics of a subject
individual may be received by obtaining the subject individual's
answers to a series of questions about the subject, such as those
illustrated in FIG. 3A. The answers may represent information known
by the subject. For example, the answers may indicate the subject,
hypothetically named Durant Laurent, does not smoke, has fine, red
hair, is male, is not concerned about hair loss, dandruff, or
baldness, and is concerned about white hair. Processing of the
accessed data and the received information may cause a diagnosis to
be generated (and optionally presented to the subject and/or
someone else), wherein the diagnosis may include one or more
characteristics known by the subject (e.g., characteristics the
subject is actually aware of and/or characteristics the subject
could become aware of readily without undergoing any sophisticated
analysis), but not previously provided by the subject. For example,
the diagnosis may indicate that Durant Laurent exhibits the
characteristics illustrated in FIG. 3B, even though none of those
characteristics was provided in the answers of FIG. 3A.
[0068] In addition to (or rather than) generating a diagnosis
including one or more characteristics "known" by the subject, the
diagnosis may include one or more characteristics not "known" by
the subject (e.g., characteristics that the subject could not
become aware of without undergoing a relatively sophisticated
analysis). For example, the diagnosis may also indicate that Durant
Laurent has a cholesterol level between 493 to 743 millimoles per
liter, tryglyceride level between 0.62 and 2.47, glycemia level
greater than 61.0 millimoles per liter, iron level greater than 90,
mean systolic arterial pressure level in the year 1996 (M_tas96)
greater than 123.6, mean dystolic arterial pressure level (M_tad)
of 79.6, above average anxiety score, and above average depression,
as illustrated in FIG. 3C, even though Mr. Laurent did not "know"
those characteristics. Note that the diagnosis of FIG. 3B and/or
FIG. 3C contains relatively large amounts of somewhat specific
information derived from a relatively small amount of somewhat
general data provided by Mr. Laurent in FIG. 3A.
[0069] In another example, another subject, hypothetically named
Yolanda Lauex, may provide answers to questions, as illustrated in
FIG. 3D, in a manner similar to that described above for Durant
Laurent. From the information provided in the answers, Yolanda
Lauex may be diagnosed with characteristics illustrated in FIGS. 3E
and 3F. FIG. 3E illustrates characteristics "known" and not
provided by Yolanda. FIG. 3F illustrates characteristics not
"known" and not provided by Yolanda. Again, note that the diagnosis
in FIG. 3E and/or FIG. 3F contains a relatively large amount of
somewhat specific information derived from a relatively small
amount of somewhat general data provided by Yolanda in FIG. 3D.
[0070] As mentioned above, the data accessed in step 102 of FIG. 1
may be data organized by an artificial intelligence engine.
"Artificial intelligence" (AI) is used herein to broadly describe
any computationally intelligent system that combines knowledge,
techniques, and methodologies. An AI engine may be any system that
is configured to apply knowledge and that can adapt itself and
learn to do better in changing environments. Thus, the AI engine
may employ any one or combination of the following computational
techniques: neural network (e.g., Kohonen map, multi-layer
perceptron, etc.), constraint program, fuzzy logic, classification,
conventional artificial intelligence, symbolic manipulation, fuzzy
set theory, evolutionary computation, cybernetics, data mining,
approximate reasoning, derivative-free optimization, decision
trees, or soft computing. Employing any computationally intelligent
technique, the AI engine may learn to adapt to unknown or changing
environment for better performance. AI engines may be implemented
or provided with a wide variety of components or systems, including
one or more of the following: central processing units,
co-processors, memories, registers, or other data processing
devices and subsystems.
[0071] The AI engine may be trained based on input such as
characteristic information, expert advice, user profile, or data
based on sensory perceptions. Using the input, the AI engine may
implement an iterative training process. Training may be based on a
wide variety of learning rules or training algorithms. For example,
the learning rules may include one or more of the following:
back-propagation, real-time recurrent learning, pattern-by-pattern
learning, supervised learning, interpolation, weighted sum,
reinforced learning, temporal difference learning, unsupervised
learning, or recording learning. As a result of the training, the
AI engine may learn to modify its behavior in response to its
environment, and obtain knowledge. Knowledge may represent any
information upon which the AI engine may determine an appropriate
response to new data or situations. Knowledge may represent, for
example, relationship information between two or more
characteristics. Knowledge may be stored in any form at any
convenient location, such as a database.
[0072] In another method in accordance with features and principles
of the present invention, an AI engine, such as that described
above, may be used to generate a profile data set from accessed
data. The profile data set may be used to generate links from the
accessed data. Or the generated profile data set may contain the
links. For example, a large quantity of data may be accessed as
described above (step 102) and processed using an AI engine to
organize the data. The AI engine may be configured to organize the
data based on a predetermined group of characteristics. The group
of characteristics may be predetermined by the AI engine, an
expert, algorithm, or other mechanism. Links may be drawn from the
organized data or the organized data may contain links as described
below.
[0073] Consistent with features and principles of the invention,
the method may include accessing data about a plurality of groups
of characteristics, as illustrated at step 402 in the flow chart of
FIG. 4. Accessing may be performed via any of mechanisms described
above for step 102 in FIG. 1.
[0074] Also consistent with the invention, the method may include
processing the accessed data, using an artificial intelligence
engine, to generate binary encoded data representing modalities of
the characteristics, as illustrated at step 404 in the flow chart
of FIG. 4. A characteristic may contain qualitative information.
Qualitative information (i.e., qualitative data) may include any
information relating to or concerning one or more qualities. The
modalities of the characteristic may be qualitative ordinal
variables of the characteristic. For example, many characteristics
are by definition limited to a finite number of values (so-called
discrete variables) that may be qualitative. Take, for example, the
characteristic of marital status, which may be single, widow,
divorced, married, or separated. This kind of characteristic is
referred to as a nominal qualitative variable, and the different
possibilities of the characteristic called modalities. When a
characteristic only has two modalities, it is referred to as a
binary characteristic. For instance, a certain characteristic may
be present or absent in an individual, (e.g., whether the
individual has ever worn an earring) thus the characteristic is a
binary characteristic. In accordance with step 404 of FIG. 4,
regardless of the number of modalities associated with a
characteristic, the characteristic may be encoded into a binary
form suitable for processing using an artificial intelligence
engine.
[0075] Forms of binary encoding may include unconstrained binary
encoding, additive binary encoding, disjunctive binary encoding,
and/or any other encoding consistent and compatible with features
and principles of the invention. Unconstrained binary encoding may
be used for characteristics with two modalities. In general, these
characteristics may be encoded with values of 0 or 1. For example,
the gender characteristic may have only two modalities, so the
first modality (e.g., male) may be encoded as 1 and the second
modality (e.g., female) may be encoded as 0.
[0076] FIG. 5 illustrates an exemplary table of modalities using
additive binary encoding for a characteristic on the frequency of
usage of a temporary hair color treatment, for example. As
illustrated at question forty-three in the survey of FIG. 2, the
characteristic may have three modalities (rare, often, not at all).
The characteristic may be binary encoded as indicated in the
exemplary table of FIG. 5. For example, if an individual filling
out the survey indicates often usage of the treatment, then the
corresponding modality two, may be binary encoded to 110. Such a
table may be extended or shortened for characteristics with
different number of modalities. For example, a characteristic with
N modalities may be encoded into a binary code with N binary
digits. If a characteristic exhibits the qualitative information
associated with the I.sup.th modality, then the first to I.sup.th
binary digits may be 1 and the (I+1).sup.th to N.sup.th binary
digits may be 0.
[0077] FIG. 6 illustrates an exemplary table of modalities using
disjunctive binary encoding. The table has four modalities with
their corresponding binary encoding. Again, the table may be
extended or shortened for characteristics with different number of
modalities. For example, a characteristic with N modalities may be
encoded into a binary code with N binary digits. If a
characteristic exhibits the qualitative information associated with
the I.sup.th modality, then the all the binary digits except for
the I.sup.th binary digit may be 0 and the I.sup.th binary digit
may be 1.
[0078] A group of characteristics with their corresponding
modalities may be encoded using one or a combination of binary
encoding routines. FIGS. 7A and 7B show an example of how a group
of characteristics may be encoded using a combination of encoding
routines. FIG. 7A shows an exemplary table with a group of
characteristics and their corresponding modalities. Gender, concern
about hair loss, concern about white hair, concern about dandruff,
concern about baldness, and smoke characteristics have modalities
encoded using unconstrained binary encoding. Hair coarseness and
natural hair color characteristics have modalities encoded using
disjunctive and additive binary encoding, respectively. Using the
information shown in FIG. 7A, FIG. 7B shows an example of a group
of characteristics in the form of binary encoded data 704 for a
person, hypothetically named John, who is male, not concerned about
hair loss, dandruff, or baldness, concerned about white hair, does
not have a smoking habit, and has fine, red hair characteristics.
Labels 702 in FIGS. 7A and 7B illustrate which of the binary digits
in the binary encoded data 704 correspond to which characteristic.
Thus, label `a` indicates the first binary digit in the binary
encoded data 704 is the encoded gender characteristic; label `b`
indicates the second binary digit in the binary encoded data 704 is
the encoded concern about hair loss characteristic; and the
remaining labels indicate which remaining characteristics
correspond to the remaining binary digits.
[0079] For purposes of the present invention, binary encoding may
be performed in any feasible format and is not necessarily
restricted to the examples described above. For example, the number
of binary digits and the encoding scheme may be modified based on
system design or requirements.
[0080] The data accessed in step 402 of FIG. 4 may comprise
characteristics of a plurality of individuals. For example, in
addition to John in the above example, the binary encoded data 704
generated at step 404 in FIG. 4 may include encoded information for
a plurality of individuals, hypothetically named John, Mary, Lisa,
. . . , Kenie, and Mike, as illustrated in FIG. 7C.
[0081] According to features and principles consistent with the
invention, the method may include processing the binary encoded
data, using an artificial intelligence engine, to generate profiles
for a profile data set, as illustrated at step 406 in the flow
chart of FIG. 4. The artificial intelligence engine may generate a
virtual grid 800 representing a profile data set of profiles as
illustrated in an exemplary manner in FIG. 8. The grid 800 may
comprise profiles 802 and connections 804 linking adjacent pairs of
profiles 802. The connections 804 may be used to determine distance
(e.g., proximity) between profiles 802. A distance of "one" between
two profiles may be defined to be when the two profiles have a path
with a single connection between them. A distance of "two" between
two profiles may be defined to be when any path of connections
between the two profiles has a minimum number of two connections.
In a similar manner, a distance of M between two profiles may be
defined to be when any path between the two profiles has a minimum
number of M connections. The connections 804 may be used to define
a set of profiles within a certain proximity of a particular,
individual profile. For example, the connections 804 may be used to
specify a neighborhood of a particular profile such as profile 806.
A neighborhood of "one" for profile 806 would comprise a set of all
profiles within one connection of profile 806. For example, all
profiles in dotted box 808 are in the neighborhood of "one" for
profile 806. All profiles in dotted box 810 are in a neighborhood
of "two" for profile 806. Analogous neighborhoods exist for the
other profiles. All or some information associated with profiles in
a neighborhood may be referred to as neighborhood data.
[0082] Configurations of profiles are not limited to
two-dimensional spaces (e.g., Cartesian plane). The profiles may be
arranged in one-dimensional space, three-dimensional space,
four-dimensional space, etc. Also, the neighborhood of a profile
may be defined using distances besides the Euclidean distance, such
as Hamming distance or any other metric known in the art and
compatible with the present invention.
[0083] Any number of profiles may be in a profile data set. The
profiles may be arranged in any configuration compatible with the
invention. Further, the neighborhoods of the profiles may be
defined in any way compatible with the invention. For example, FIG.
9 illustrates an exemplary arrangement 900 of profiles wherein the
profiles 902 are situated on a Cartesian plane 904 with
corresponding locations defined by Cartesian coordinates. The
Cartesian coordinates of a profile in this example are defined to
be the coordinates of the center of the profile. In this
arrangement 900, a neighborhood of a profile may be defined to be
all profiles whose Cartesian coordinates are within a given
Euclidean distance, d, of the profile's Cartesian coordinates. For
example, all profiles with Cartesian coordinates within the dotted
circle 908 of radius d are within the neighborhood of d of profile
906.
[0084] All profiles may contain a referent. The referent may be a
binary code and may define the characteristics for its respective
profile using any of the binary encoding methods previously
described. For example, the referent for profile 806 in FIG. 8 may
be 101000010001000000 (the example represented in FIG. 7b).
According to the exemplary table in FIG. 7A, this referent would
indicate a profile with male gender, unconcerned about hair loss,
concerned about white hair, unconcerned about dandruff, unconcerned
about baldness, no smoking habit, and fine, red hair. The referent
for each profile may be determined during the processing step 406
of FIG. 4. This may involve using a neural clustering algorithm
such as a dynamic cluster method, a k-means clustering algorithm, a
hierarchical clustering algorithm, a mobile center method, and/or a
topology map method. (Other algorithms may also be used.) As
described in more detail later, the neural clustering algorithm may
use neighborhood data.
[0085] Neural clustering algorithms may include methods described
by Thiria, S., Lechevallier, Y., Gascuel O., and Canu, S. in
Statistique et methodes neuronales, Dunod, Paris, 1997, which is
incorporated herein by reference in its entirety.
[0086] A dynamic cluster method may begin by initializing the
referent for each profile. The referent may be initialized
randomly, pseudo-randomly, using an algorithm, and/or using any
technique compatible with the invention. After initialization, the
referent may be trained using binary encoded data representing
characteristics from a plurality of groups. In the above example,
FIG. 7C illustrates a plurality of binary encoded data for a
plurality of respective individuals. Each one of the plurality of
binary encoded data may be assigned to a profile with the closest
matching referent. The profile with the closest matching referent
may be determined using a metric that compares a degree the
referent matches the binary encoded data of an individual. One
metric, known as the hamming distance, inter alia, may be defined
as
d ( z i , p ) = j = 1 B z i ( j ) - r p ( j ) , ##EQU00001##
wherein d(z.sub.i, p) is the degree the referent r.sub.p of profile
p matches the binary encoded data z.sub.i of the i.sup.th
individual, z.sub.i(j) is the j.sup.th binary digit in z.sub.i,
r.sub.p(j) is the j.sup.th binary digit in r.sub.p, and B is the
number of binary digits in the binary encoded data. For this
metric, a degree of lower value indicates a closer match. For
example, the degree of match between binary encoded data
101000010001000000 and referent 101100010000100000 would be three
and the degree of match between binary encoded data
101100010000100000 and referent 101100010000100000 would be zero.
Binary encoded data that matches the referents of more than one
profile with equal degree may be arbitrarily assigned to one of the
profiles. A profile may have more than one binary encoded data
assigned to it.
[0087] Once all the binary encoded data from the plurality of
individuals are assigned to a profile, the referent of each profile
may be optimized. Optimization for the referent of a given profile
may be performed by calculating a center median of all the binary
encoded data assigned to profiles within a neighborhood of the
given profile and then replacing the referent of the given profile
with the center median. More particularly, the referent for the
given profile may be optimized by
r c ( j ) = median z i .di-elect cons. N c ( z i ( j ) ) ,
##EQU00002##
wherein r.sub.c(j) is the j.sup.th binary digit of the optimized
referent r.sub.c for the given profile c, and N.sub.c is the set
containing all binary encoded data z.sub.i assigned to profiles
within a neighborhood of the given profile c. In the above example,
if profile 806 in FIG. 8 is assigned binary encoded data z.sub.5
and z.sub.7, profile 812 is assigned binary encoded data z.sub.4
and Z.sub.13, and profile 814 is assigned binary encoded data
z.sub.6, z.sub.10, and z.sub.21, then the optimized referent r for
profile 806 using binary encoded data within a neighborhood of one
may be determined as illustrated in the table of FIG. 10. The
j.sup.th binary digit of the referent r in profile 806 is the
median of the j.sup.th binary digit of all the binary encoded data
in the neighborhood (i.e. z.sub.4, z.sub.5, z.sub.6, z.sub.7,
z.sub.10, z.sub.13, and z.sub.21).
[0088] In optimizing the referent of a given profile, the binary
encoded data may be weighted as function of the proximity between
the given profile and the profile to which the binary encoded data
are assigned. FIG. 10 illustrates a center median calculated using
uniform weighting because the binary encoded data used to determine
the center median were all weighted with one. However, the center
median in FIG. 10 was only calculated using binary encoded data
from profiles within a neighborhood of one, so binary encoded data
from profiles outside the neighborhood of one may be viewed as
having a weight of zero. This may be characterized by the indicator
function in FIG. 11, wherein the x-axis represents the distance of
a neighboring profile from the given profile and the y-axis
represents the weight applied to the binary encoded data in the
neighboring profile. Other forms of weighting may include the
Gaussian function illustrated in FIG. 12 or any other function
compatible with features and principles of the present invention.
Example of other functions may include, but are not limited to,
discrete functions, continuous functions, probabilistic functions,
etc.
[0089] After optimizing the referents of all the profiles, the
binary encoded data in the profiles may be re-assigned to the
profiles with the closest matching optimized referent. Referents of
all the profiles may be optimized again using the re-assigned
binary encoded data in the manner described above. The assignment
of the binary encoded data and optimization of the referents may be
repeated for a predetermined number iterations or until the
referents converge (i.e., no longer change from one iteration to
the next). The weighting applied to the binary encoded data during
optimization may change for each iteration. For example, the
weighting may decrease the number of binary encoded data used to
optimize a referent by assigning a weight of zero to profiles
beyond a certain distance from the profile having its referent
optimized and decreasing the distance at each iteration. Decreasing
this distance effectively shrinks the neighborhood of profiles used
to optimize a referent.
[0090] The dynamic cluster method is finished once the assignment
of binary encoded data and optimization of the referents are
complete. FIG. 13 illustrates an example where the binary encoded
data for individuals may be assigned in grid 800 of FIG. 8 after
the dynamic cluster method is completed. For example, profile 1302
in grid 1300 has the binary encoded data of Mac and Kenie assigned
to it because the profile 1302 contains the referent with the
closest match to their binary encoded data. Further, profiles that
are within proximity of each other may contain referents that are
similar to each other because neighborhoods are used with the
dynamic cluster method in optimizing referents. Therefore, profiles
1304 and 1306 may contain referents similar to the referent in
profile 1302 and all the binary encoded data (i.e., Mike, Sean,
Lewis, Greg) assigned to profiles 1304 and 1306 may be similar to
the binary encoded data (i.e., Mac and Kenie) assigned to profile
1302. As a result, Mac and Kenie may have characteristics similar
to characteristics exhibited by Mike, Sean, Lewis, and Greg.
[0091] The above description of dynamic clustering may use the
center median and/or Hamming distance for determining and
optimizing the referent of each profile. Alternatively, the mobile
center method may also be used for the same purpose. Instead of the
center median and/or Hamming distance, the mobile center method may
use a center of gravity (i.e., center mean) and/or Euclidean
distance.
[0092] The above description for generating the clustered profiles
in FIG. 13 uses a dynamic cluster method, but other methods
compatible with the present invention may also be used. For
instance, a k-means clustering algorithm may be used instead or in
combination with the dynamic cluster method. Unlike the dynamic
cluster method, which may optimize the referent of each profile
when all the binary encoded data are assigned to the profiles, the
k-means clustering algorithm may optimize the referent of a profile
when less than all the binary encoded data have been assigned to
the profiles. The k-means clustering algorithm may optimize the
referent of each profile after each assignment of binary encoded
data to a given profile. In the k-means clustering algorithm, the
selection of at least one binary encoded data to be re-assigned may
be performed randomly, pseudo-randomly, or using a prescribed
method.
[0093] For example, at least one, but not all, of the binary
encoded data assigned to the profiles may be selected and
re-assigned to the profiles. After re-assigning, the referents of
all of the profiles may be optimized. The referents may be
optimized by calculating the center median of the binary encoded
data in each profile. The assignment of at least one, but not
necessarily all of, the binary encoded data and optimization of the
referents may be repeated for a predetermined number iterations or
until the referents converge (i.e. no longer change from one
iteration to the next).
[0094] Alternatively, the optimization of the referents may use the
dynamic cluster method at one iteration and the k-means clustering
algorithm at another iteration. For example, at one iteration all
the binary encoded data may be re-assigned and then the referents
may be optimized. At another iteration, at least one, but not
necessarily all of, the binary encoded data may be re-assigned and
then some or all the referents may be optimized.
[0095] A hierarchical clustering algorithm may be applied to the
dynamic clustered profiles in FIG. 13 (or k-means-clustered
profiles or combined dynamic clustered and k-means-clustered
profiles). The hierarchical clustering algorithm may reduce the
number of profiles in grid 1300 by agglomerating the profiles in
grid 1300. The first profile may be merged with a second profile,
wherein a clustering index indicates the referent of the first
profile matches closer to the referent of the second profile than a
referent in any other profile in grid 1300. The clustering index
may be a Ward index, pseudo-Ward index, a distance index, an index
taking total number of binary encoded data, or any other index.
[0096] The Ward index is defined to be
.DELTA. Ward = n a n b n a + n b j = 1 B r a ( j ) - r b ( j ) ,
##EQU00003##
wherein n.sub.a is the number of binary encoded data assigned to
profile a, n.sub.b is the number of binary encoded data assigned to
profile b, r.sub.a is the referent of profile a, r.sub.b is the
referent of profile b, B is the number of binary digits in the
referent, r.sub.a(j) is the j.sup.th binary digit in referent
r.sub.a, and r.sub.b(j) is the j.sup.th binary digit in referent
r.sub.b. For example, the referents of profiles 1302 and 1304 in
FIG. 13 may be 100000001001111111 and 100000010001111110,
respectively r.sub.a and r.sub.b. The numbers of binary encode data
in profiles 1302 and 1304 are two and one, respectively n.sub.a and
n.sub.b. Therefore, the Ward index between the two profiles is
calculated to be two.
[0097] The pseudo-Ward index is defined to be
.DELTA..sub.Pseudo(P.sub.a,P.sub.b)=.PHI.(P.sub.a.orgate.P.sub.b)-.PHI.(-
P.sub.a)-.PHI.(P.sub.b),
wherein P.sub.a and P.sub.b are sets of binary encoded data
contained in the profiles a and b, respectively, and .PHI.(S) is
the inertia of a given set S. For example, .PHI.(P.sub.a) is
defined to be
.PHI. ( P a ) = i = 1 n a j = 1 B z a , i ( j ) - r a ( j ) ,
##EQU00004##
wherein z.sub.a,i is the i.sup.th binary encoded data in profile a,
and z.sub.a,i(j) is the j.sup.th binary digit in z.sub.a,i.
[0098] The distance index is defined to be
.DELTA. Dist = j = 1 B r a ( j ) - r b ( j ) . ##EQU00005##
In the above example, the distance index between profiles 1302 and
1304 would be three.
[0099] The index taking total number of binary encoded data is
defined to be
.DELTA. Total ( P a , P b ) = 2 .PHI. ( P a P b ) B ( n a + n b ) .
##EQU00006##
[0100] The referent for the merged profile of two merging profiles
may be set to the referent of the merging profile with more binary
encoded data assigned to it. For example, if profiles 1302 and 1304
are merged, the referent of the merged profile may be the referent
of profile 1304 because profile 1304 has more binary encoded data
assigned to it than profile 1302. Alternatively, the referent of
the merged profile may be set using any other criteria or algorithm
compatible with the invention. Profiles in the grid may continue to
be merged together until only a predetermined number of profiles
remain, until the combined profiles contain a certain number of
binary encoded data, or until any other criteria for final
clustering is achieved.
[0101] After hierarchical clustering, an exemplary set of merged
profiles may look like the example shown in FIG. 14. Since the
clustering algorithm described above may organize individuals in
the grid 1400 by placing individuals with similar binary encoded
data in the same profile, the individuals in a given profile may
have similar characteristics. Further, the referent of any profile
may be binary data and can be decoded using the table in FIG. 7A to
a corresponding referent group of characteristics. The referent
group of characteristics of a profile may be substantially
representative of the characteristics exhibited by individuals in
the profile. For example, the referent of profile 1402 may be
011001001001111100. Using FIG. 7A, the group of characteristics
corresponding to the referent is female gender, concerned about
hair loss, white hair, and baldness, not concerned about dandruff,
smoking habit, and average, dark chestnut blonde hair. Therefore,
all individuals in profile 1402 may substantially exhibit the
characteristics of the referent group.
[0102] According to features and principles of the present
invention, as illustrated at step 408 in the flow chart of FIG. 4,
the method may include assigning at least some of the plurality of
groups of characteristics, the accessed data, and/or the binary
encoded data using the artificial intelligence engine, to the
profiles to generate a profile data set. With reference to the
example shown in FIG. 14 and as described above, the binary encoded
data assigned to the profiles in the figure reflect a group of
characteristics (e.g., gender, concern about hair loss, concern
about white hair, concern about dandruff, concern about baldness,
smoking habits, coarseness of hair, and color of hair) for a
plurality of individuals. Additional groups of characteristics of
the individuals not used to organize the individuals in grid 1400
may be assigned to the profiles containing the individual. For
example, an additional group of characteristics of individuals in a
profile, including characteristics such as whether there is
occurrence of hair loss and usage of fingernail multivitamin
treatment, may be assigned to the profile containing the respective
individual. If the individuals in a profile substantially exhibit
the same characteristics in the additional group of
characteristics, then a link may be made between the
characteristics reflected in the referent of the profile and the
additional group of characteristics. However, a link may also be
made within the characteristics reflected in a referent of the
profile by virtue of the fact that the generation of the profiles
at step 406 uses binary encoded data reflecting characteristics to
generate the profiles.
[0103] For example, if individuals in profile 1402 in FIG. 14
substantially exhibit hair loss and frequent usage of fingernail
multivitamin treatment, then a link may be made indicating a
subject individual exhibiting characteristics reflected in the
referent of profile 1402 has likely exhibited in the past, is
likely exhibiting at the present time, or will likely exhibit in
the future, hair loss and frequent usage of fingernail multivitamin
treatment.
[0104] Once the profile data set is generated it may be updated.
Additional data from other individuals may be collected to train
the referents in the profile data set. The additional data may be
added to previous characteristic data and processed as described
above to generate a new profile data set.
[0105] With reference to FIG. 1 and FIG. 4, the diagnostic method
as illustrated in the flow chart of FIG. 1 may use a profile data
set, for example to diagnose a subject individual. In such an
example, at step 102 in FIG. 1, the accessed data about the
plurality of groups of characteristics may include a first group of
characteristics from a plurality of individuals. The first group
may comprise gender, concern about hair loss, concern about white
hair, concern about dandruff, smoking habit, concern about
baldness, coarseness of hair, and color of hair characteristics. At
step 106 in FIG. 1, the group of characteristics from the plurality
of individuals may be used to organize the individuals into a
profile data set. The plurality of individuals may be assigned to
profiles in the profile data set based on their respective
characteristics from the first group or other characteristics in
the plurality of characteristics. The individuals may be assigned
to profiles according to how close their characteristics match the
characteristics reflected in the referent of each profile.
[0106] Further, the received information at step 104 in FIG. 1 may
reflect the exhibition of the first group of characteristics of a
subject individual. The received information may be binary encoded
or already binary encoded. The profile with the referent matching
closest to the binary encoded data may contain individuals
displaying characteristics substantially similar to the subject
individual. The profile may also contain additional characteristics
(e.g., a second group of characteristics) not used to organize the
individuals in the profile data set. Therefore, the subject
individual may be diagnosed to have a predisposition to exhibit the
additional characteristics of the individuals in the profile.
[0107] A third embodiment of the invention may include a dynamic
surveying method. According to features and principles of the
present invention, the method may include accessing data organized
by an AI engine as illustrated at step 1502 in the flow chart of
FIG. 15. Accessing may be performed using mechanisms as previously
described. For example, the accessed data may include personal
information such as information identifying characteristics and
information for predicting evolution of the identified
characteristics. Information for predicting the evolution of the
identified characteristics may include links derived using the
artificial intelligence as previously described. The accessed data
may include a profile data set generated as previously
described.
[0108] Consistent with the present invention, the dynamic surveying
method may include accessing queries and presenting to a subject a
subset of queries from the accessed queries, as illustrated at
steps 1504 and 1506 in the flow chart of FIG. 15. The accessed
queries may be stored in some form using any storage medium
described above or generated. The queries may include questions
that elicit one or more answers that may be used to process at
least some of the data accessed in step 1502. For example, one or
more answers to the queries may reflect whether a group of
characteristics are exhibited by the individual. Additionally (or
alternatively), the group of characteristics may be different from
the one or more characteristics being diagnosed and may be linked
with the characteristics being diagnosed. Presenting of the query
subset to the subject may include direct or indirect actions.
Direct actions of presenting may include displaying queries on a
video screen, orally posing the queries to the subject, providing a
printed document containing the queries, etc. Indirect actions may
include preparing queries for subsequent actions directly
presenting the queries, such as sending queries to a third party to
pose the queries to the subject, converting queries into digital
form for presenting over a network, etc.
[0109] Consistent with the present invention, the dynamic survey
method may include selecting, for at least some of the queries
presented, a next query as a function of the subject's answer to a
previous query, as illustrated at step 1508 in the flow chart of
FIG. 15. The selecting could be performed using an algorithm, an
artificial intelligence engine, and/or any other techniques.
Selecting may include determining queries that elicit answers
reflecting whether the subject exhibits a first group of
characteristics. The queries may be selected to present the fewest
number of queries to the subject before providing at least one of a
diagnostic, advice, and/or information to the subject.
Alternatively (or additionally), the queries may be selected
linearly or sequentially until the subject's answers to the
selected queries supply enough information to provide at least one
of a diagnostic, advice, and/or information to the subject. A next
query in a series of queries may be selected based on the answer to
a previous query.
[0110] For example, one embodiment of the present invention may
involve accessing a profile data set, such as one illustrated in
FIG. 13 or 14. The profile data set may be generated using
information from a survey (or other mechanism), such as the
exemplary survey illustrated in FIG. 2, as previously described.
Queries used to determine which characteristics are exhibited by a
subject may be accessed and presented to the subject. The answers
to the queries may be used to determine which profile in the
profile data set best describes or is most suitable to the subject.
For example, the queries may ask the subject whether the subject is
a man or a woman, whether the subject is concerned about hair loss,
white hair, or dandruff, whether the subject smokes, whether the
subject has certain degrees of oiliness of hair, etc. A profile in
the profile data set that most closely resembles answers provided
by the subject may be selected. Additional information (e.g., Body
Mass Index (BMI), blood characteristics, etc.) in the selected
profile and not provided in the answers of the subject may be
presented to the subject (or to another entity, such as a
practitioner) as a diagnosis. In addition (or in the alternative),
certain embodiments of the method may involve generating advice
and/or information based on the processing of the accessed
data.
[0111] In selecting the profile that most closely resembles answers
provided by the subject, portions of the profile data set may be
narrowed (e.g., reduced, excluded, etc.) until a suitable profile
remains for presenting to the subject. Prior to finding a suitable
remaining profile, the queries may also be used to indicate whether
the subject resembles at least some characteristics and/or aspects
in a particular profile in the profile data set. The queries may be
selected to efficiently exclude portions of the profile data set.
For example, if half of the profiles in a profile data set have
male gender for a gender characteristic and half have female
gender, and if a third of the profiles have oily hair, a third have
non-oily hair, and a third have very oily hair, then a query asking
whether the subject has non-oily, oily, or very oily hair would
narrow a portion of the profile data set more efficiently than a
query asking whether the subject is male or female. More
particularly, once a subject answers a query about oiliness of
hair, two-thirds of the profiles in the profile data set may be
excluded because those profiles do not apply to the subject.
Whereas, once a subject answers a query about gender, the profile
data set is merely reduced by one-half. According to such an
example, one embodiment of the method includes presenting a query
relating to oiliness of hair before considering presenting a query
relating to gender.
[0112] As one of ordinary skill in the art will appreciate, various
algorithms (e.g., tree-searching algorithms, shortest-path
algorithms, network algorithms, etc.) for selecting queries to
efficiently exclude profiles of a profile data set based on the
answers from the subject and characteristics contained in the
profiles may be used with the present invention. Further, besides
algorithms, various artificial intelligence engines, as described
above, may be used to select the queries to efficiently reduce the
profile data set.
[0113] Alternatively, the queries may be selected sequentially
through a sequence of queries about characteristics in the
profiles. For example, a list of queries asking the subject about
gender, concern about hair loss, concern about white hair, concern
about dandruff, smoking habits, etc. may be presented to the
subject sequentially or in a prescribed order regardless of how
efficient the queries and answers are in narrowing portions of the
profile data set. Once the subject answers enough queries to
identify a profile in the profile data set that best matches the
characteristics exhibited by the subject, then no additional
queries may need to be presented to the subject.
[0114] As mentioned above, whether the queries are selected
efficiently or inefficiently, according to an exemplary embodiment
at least one next query selected and presented to the subject may
be based on at least one answer to a previous query. For instance,
in the previous example, a profile data set may contain an equal
number of profiles with non-oily, oily, and very oily hair
characteristics. However, all the profiles with the non-oily hair
characteristic may contain the concerned about dry hair
characteristic, all the profiles with the oily hair characteristic
may contain the not concerned about dry hair characteristic, half
the profiles with the very oily hair characteristic may contain the
concerned about dry hair characteristic and the remaining half of
the profiles with the very oily hair characteristic may contain the
not concerned about dry hair characteristic.
[0115] If the subject's answer to a previous query about oiliness
of hair is that the subject exhibits the non-oily hair
characteristic, then a next query may not ask about the subject's
concern about dry hair characteristic because the answer to such a
query either won't further narrow the reduced profile data set
containing the oily hair characteristic or will completely exclude
all profiles in the reduced profile data set containing the oily
hair characteristic. Instead, the next query may be selected to
more effectively narrow portions of the reduced profile data set
containing the non-oily hair characteristic. If the subject's
answer to a previous query about oiliness of hair is that the
subject has very oily hair, then the next query may ask the subject
whether the subject is concerned about dry hair.
[0116] Features and principles of the present invention may be
implemented in a system comprising a data processor (or data
processors) and a storage medium (or storage media). The data
processor and storage medium may be functionally coupled. The
storage medium may contain instructions to be executed by the data
processor for performing methods consistent with the present
invention. Data processors may include desktops, mainframes,
computers, application specific integrated circuits, electronic
devices, mechanical devices, and/or any other mechanism for
executing instructions. The storage medium may include any storage
media as previously described.
[0117] For example, features and principles of the present
invention may be implemented in system 1600 of FIG. 16. The system
1600 may comprise a computer kiosk 1602, a computer terminal 1604,
a network 1606, a mainframe 1608, and a database 1610. A plurality
of individuals may fill out a survey containing data about
characteristics of the individuals at the computer kiosk 1602. The
mainframe 1608 may access the data via the network 1606. A subject
individual may send information via the network 1606, in response
to a series of queries, reflecting that the subject exhibits some
of the characteristics. The queries may be selected using the
dynamic surveying method described above. The mainframe 1608 may
receive the information and process the received information and
accessed data to generate a diagnosis, advice, and/or any other
information. The mainframe 1608 may be configured to generate a
profile data set by accessing data, processing the accessed data to
produce binary encoded data, processing the binary encoded data
using an artificial intelligence engine to generate profiles, and
assigning the characteristics to the profiles to generate the
profile data set.
[0118] The computer kiosk 1602, computer terminal 1604, and
mainframe 1608 may be any computing platform, including a
stand-alone personal computer or networked computers.
Alternatively, the system may be implemented in a single computer
or a group of computers in a LAN or WAN.
[0119] As used herein, the words "may" and "may be" are to be
interpreted in an open-ended, non-restrictive manner. At minimum,
"may" and "may be" are to be interpreted as definitively including
structure or acts recited.
[0120] Further, while flow charts presented herein illustrate a
series of sequential blocks for exemplary purposes, the order of
blocks is not critical to the invention in its broadest sense.
Additionally, blocks may be omitted and others added without
departing from the spirit of the invention. The invention may
include combinations of features described in connection with
differing embodiments.
[0121] For example, in the exemplary method, illustrated in flow
chart 400 of FIG. 4, for generating a profile data set, step 408
may be removed. The profiles generated at step 406 may together
form a profile data set. This profile data set may be used with a
diagnostic method, dynamic surveying method, or other features and
principles consistent with the present invention.
[0122] In another example, the profiles generated at step 406 may
not include using a hierarchical clustering algorithm. The
profiles, as illustrated in FIG. 13, generated before agglomeration
by the hierarchical clustering algorithm may together form another
profile data set. This profile data set may also be used with a
diagnostic method, dynamic surveying method, or other features and
principles consistent with the present invention.
[0123] In the foregoing description, various features are grouped
together in various embodiments for purposes of streamlining the
disclosure. This method of disclosure is not to be interpreted as
reflecting an intention that the claimed invention requires more
features than are expressly recited in each claim. Rather, as the
following claims reflect, inventive aspects may lie in less than
all features of a particular embodiment described above. Thus, the
following claims are hereby incorporated into this description,
with each claim standing on its own as a separate embodiment of the
invention.
* * * * *