Transcript Oracle Data Mining for Java (DM4J) Applied to Epidemiological

Oracle Data Mining and
Epidemiological Analysis
Scott A. Rappoport, OCP
MTS Technologies
OracleWorld 2003
San Francisco, CA
paper #63144
Presentation Goals
 Short intros
 Vocabulary


Present Basic Medical Terms
Describe Data Mining Models and Terms
 Synthesize


What questions are we asking?
Applying DM to Epidemiological issues
 Demonstrate the DM4J components
 The future:


Challenges
10g features
paper #63144
The DM Dimension
 Data Mining capability readily accessible to the end
users opens a whole new dimension of what can be
performed in the medicine.
 New questions are being generated based on the
availability of these new techniques.
 This is a cutting edge (bleeding edge) advanced
technique
paper #63144
A few disclaimers…
Medical data is highly sensitive information… Thus:
 No personally identifiable info is presented
 No specific aggregated information on disease
types, locations, or time is provided
 Scaled back list of attributes in demos
 However, demos will give an indicative application
of the technology.
paper #63144
About you
 What percentage of the audience:

Has a medical background?



Has an IT background?



Physician
Epidemiology/research/academic
DBA
Developer
Knows a lot about Data Mining? Statistics?
 Has at least two of the above? Three?
paper #63144
About me
 Oracle Certified DBA and Developer
 ASQ Certified Quality Engineer
 Principal Architect, supporting the Naval Health
Research Center in San Diego, CA
 Instructor of Oracle, Data Warehouse, and Web
Services courses at UCSD-Extension
 Papers on Java, DataWarehousing – IOUG/ODTUG
 Biochemistry degree/ worked in a diagnostics firm
 Son of a clinical pathologist
paper #63144
Let’s get at it…
The Medical Side
paper #63144
Medical Lexicon
 Epidemiology

Study of the relationships of various factors determining the
frequency and outbreak of disease.
 Nosocomial

Outbreaks originating within a hospital.
 Nosology

Study of the classification of diseases.
 ICD9/10

International Classification of Diseases: v9 or 10. Classification of
disease by major category – represented by a three-digit code,
followed by a specific type, represented by a two-digit code.
 DNBI:

Disease Non-Battle Injury. Military classification of disease types.
paper #63144
Nosology/ICD9 Disease Classification
 Over 12,000 separate diseases
 Classified into 13 areas
 Further sub-classed
 Set off by 3 digit code, then additional 2 digit
descriptor for better granularity
 DNBI – military designations
paper #63144
Epidemiological/Medical Practice
Questions
 What factors affect the onset of disease within a
population?
 What is the likelihood that a patient will require
follow-up treatment, hospitalization, or that the case
will worsen?
 Are there particular clusters of patients that are
more likely to develop a certain disease?
 How often is a case mis-diagnosed?
 Is a particular treatment likely to cure the ailment?
paper #63144
Summarizing the Concerns
 Predictive concerns
 Classification of risks and subjects
 Attribute ranking concerns
 Multi-factor relevance
 Dealing with large numbers of attributes
 Clustering questions
 Unknown associations
paper #63144
Epidemiological techniques
 Statistical packages



Chi-square
ANOVA / ANCOVA / MANOVA
Multi-variate Analysis (Attribute Scoring):
Multiple Logistic Regression (binomial/dichotomous)
 Multiple Linear Regression (multiple/category)



Covariance
2x2 matrix
paper #63144
Risk factors/classification








Environmental: exposure, location, job risks, diet
Genetic: Genetic markers present?
Clinical: Blood/other diagnostics data
Familial: Other family members? Who, what?
History: Past illnesses? What? When? How often?
Socio-economic: Job, married, education, age, gender
Lifestyle: Exercise, smoker, alcohol
Ethnic/National/Geographical
paper #63144
Patient Data Universe
Geographic
factors
Treatment Fac/
Personnel
Patient history
Physician's
note
Family History
Total Patient Description
Diagnostic
data
Drug
interactions
Genomic
paper #63144
Ethnic/race/
national
Lifestyle
factors
A vast amount of data potentially to be collected and
mined in the patient data universe !!!
The Data Mining Side
paper #63144
Reporting techniques/hierarchies
User Sophistication
Data
Mining
What hidden
associations or
clusters of attributes
may exist?
On-Line
Analytical
Processing
(OLAP)
Ad Hoc
Queries
Operational Reporting
Data
paper #63144
What is likely to
happen tomorrow
(based on past trends/
aggregations)?
Why did that happen
yesterday?
What (specific events)
happened yesterday?
Reporting Examples
Query
Technique
Operational
reporting
Reporting needs
Example
Basic information on an event
Find the diagnosis of patient
#A1234 on this date.
Ad-hoc
User define queries to help understand
an event
Does the specific patient have a
past history of such a
diagnosis?
OLAP
Summarized data of events across many
dimensions
What is the incidence rate of
this disease among this patient
type? For this area, season,
hospital, etc? Is this becoming
more prevalent?
Data Mining
Attribute associations, predictive
modeling, clustering of populations by
attribute sets.
Across many attributes and records
What are the risk factors for
this disease? What is the
likelihood a treatment will
succeed for a patient? What
specific populations are at risk?
paper #63144
Data Mining Techniques
 Classification

Seeks to find out attributes that best predict a dependent variable
 Clustering

Seeks groupings of attributes in populations
 Association

What is the likelihood that event A will lead to or occur with event
B, C, or D…
 Attribute Importance

Ranking of attributes based on their effects on a given dependent
variable
 Lift Model:

Measures how well a model can identify a given target
paper #63144
Data Mining Terms
 Confusion Matrix:

Tests model accuracy. Actual to predicted evaluated, scored by
incidence of false-positives / false-negatives.
 False-negative:

disease present, results not shown
 False-positive:

disease not present, results show
 Supervised learning:

target value is specified. Classification / regression
 Unsupervised learning:

Relations/target attributes not known. Clusters/Assoc
paper #63144
Data Mining Terms (cont’d)
 Support:

The measure of how often the collection of items in an
association occur together as a percentage of all the
transactions.
 Confidence:

Confidence of rule "B given A" is a measure of how much
more likely it is that B occurs when A has occurred.
 ROC:

Receiver Operating Characteristic. Used in Lift models to
determine how well the model identifies targets as opposed
to random selection.
paper #63144
Supervised/Unsupervised
Supervised  Prediction  odds of success

Classification
Model
 Test (obtain false-positives/negatives
 Apply
 Lift


Attribute Importance
Determine attributes with the most effect on result
 Want to split on this attribute

paper #63144
Supervised/Unsupervised
 Unsupervised  No a priori knowledge 
find hidden relations/ associations/ groupings

Clustering


What groups of subjects share values of attributes that
are closely related?
Associations

paper #63144
Find events that are related; i.e., if A (and/or B)
happens, what are the odds that C will happen?
Classification Modeling
 Used to find a predictive model of independent
attributes on the outcome of a dependent attribute
 Algorithms: Naïve Bayes, Adaptive Bayes NetWork
Attributes
Branches
Pruned
.
.
.
.
.
paper #63144
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Number
of Levels
Classification Model (cont’d)
 Replaces:

Multi-variate Analysis


Multiple Logistic Regression (binomial/dichotomous)
Multiple Linear Regression (multiple/category)
 Questions:



Given a set of factors, what is the likelihood that a
disease will be expressed?
What is the likelihood the disease will lead to a more
severe ailment?
What category (multi-option) of health based on inputs?
paper #63144
Classification Model: To Do’s
1. Create a model: Classification Build
2. Refine: Run an Attribute Importance Model
to help define best attributes to “split”
3. Test the model: Classification Test
4. Predict results: Classification Apply
5. Targeting: Classification Lift
paper #63144
Clustering
 Unsupervised model that attempts to find groups
within the population that share similar attributes
 Algorithm: k-means, O-Cluster
Age
AGE
INCOME
C2
Rank
C2
INCOME
C1
Age
AGE
C1
C1
Rank
INCOME
paper #63144
Centroids
AGE
Histograms
Courtesy Charlie Berger, Oracle
Clustering (cont’d)
 k-means only takes numeric values, and requires
the number of clusters to be specified. Good for
smaller datasets with fewer attributes.
 O-Clusters: more robust than k-means
 Questions:



What groups of people are present in a population, and
what are their common attributes?
How are the members distributed along those attributes?
Are there given clusters of people related to a specific
disease family? Are members more or less susceptible?
paper #63144
Association Models
 Unsupervised model that returns a set of rules
determining if one or more attributes are associated
with other attributes.
 Scored by support/confidence
 What is the likelihood of A happening if B
happens?
 Often used with sparsely populated data sets.
 Questions:

What is the relationship between overweight recruits,
smoking, and attrition in boot camp?
paper #63144
Applications/Demos
 Review of the parts of the process:


JDeveloper9i layout, model wizards, creation, run
ODM Browser: task review, navigation, results
 Creation of models in JDeveloper9i with DM4J
Wizards




Clustering Model Build and analyze histograms
Association Model Build: Analyze rules
Classification Model: Build, Test, Apply, Lift
Attribute Importance
paper #63144
Challenges
 Most data sources have not been modeled to collect
the range of data needed.
 Bio-informatics opens a whole new range of study
not even imagined a few years ago.
 Data Stores are inconsistent.
 Doctors notes are not uniform.
 Legacy Apps are a mess. (COBOL, poorly
documented, personnel retired…)
paper #63144
More challenges
 Vast amounts of data/ processing
 Confusion matrix on attributes with large
categories.
 Structuring questions “to peel away” masking
factors, and be sensitive to subtle associations
 Bringing it to the masses
 Overcoming resistance to change.
paper #63144
New Native
g
10
Features
 Text Mining – to help us search through physicians’
notes
 Support Vector Machines (SVM): “Neural
Networks on Steroids.”
 Non-negative Matrix Factorization (NMF):
Algorithm to help “boil down” many attributes into
a manageable set.
 Enhanced Bio-informatics support in the DB.
 Transformation creation (currently alpha)
paper #63144
Summary
 Covered a multi-disciplinary topic
 Attempted to show how DM is uniquely suited to
Epidemiological study
 Showed the ease by which models can be made
 Still, model creation requires trained personnel
 Many challenges remain to fully exploit this
technology.
paper #63144
Questions?
paper #63144
Special Thanks to….
 Mark Kelly, Oracle Data Mining
 Robert Haberstoh, Oracle Data Mining
 Charlie Berger, Director Oracle Data Mining
paper #63144
Follow-up
 Please fill out the on-line survey
 Session #63144
 Feel free to contact me:
Scott Rappoport, OCP
Principal Technical Staff Member
MTS Technologies
619-725-5082
[email protected]
paper #63144