Pres Db Data Mining 10gr2 0806

8/8/2019 Pres Db Data Mining 10gr2 0806

1/61

This presentation is for informational purposes only and may not be incorporated into a contract or agreement.


2/61

The following is intended to outline our general product direction. Itis intended for information purposes only, and may not be

incorporated into any contract. It is not a commitment to deliver anymaterial, code, or functionality, and should not be relied upon in

making purchasing decision. The development, release, and timingof any features or functionality described for Oracles products

remains at the sole discretion of Oracle.


3/61

Copyright 2006 Oracle Corporation


4/61


Data Warehousing

ETL

OLAP

Data Mining

Oracle 10Oracle 10 g g DBDB

Statistics

Oracle In-Database Advanced AnalyticsStatistics, Data Mining, Text Mining, & More!

Charlie BergerSr. Dir. Product Management, Life & Health Sciences Industry & Data Mining TechnologiesOracle [email protected]
mailto:[email protected]:[email protected]


5/61Copyright 2006 Oracle Corporation

Agenda

Oracle Data Mining Overview Demos

Oracle Data Mining Integration with Oracle BI EE Spreadsheet Add-in for Predictive Analytics Text Mining

Code Generation Release In-Database Analytics Example

Comparison to SAS Partners Summary

Data Warehousing

ETL

OLAP

Data Mining


Statistics



The Evolving Role of BI

Analysts

Historical data

Reporting results

From: To:

Pervasive use

Real-time, predictive data

Insight-driven business

process optimization

Unified, enterprise viewFragmented view

Unified infrastructure &prebuilt analytic solutions

Analytic tools


7/61



What is Data Mining?

Process of sifting through massive amountsof data to find hidden patterns and discovernew insights

Data Mining can provide valuable results: Identify factors more associated with a target

attribute (Attribute Importance) Predict individual behavior (Classification) Find profiles of targeted people or items

(Decision Trees)

Segment a population (Clustering) Determine important relationships with the

population (Associations) Find fraud or rare events (Anomaly Detection)


9/61


Business Intelligence & Analytics

Knowledge discoveryof hidden patterns

Who will buy a mutualfund in the next 6months and why?

Extraction ofdetailed androll up data

Who purchasedmutual funds inthe last 3 years?

Summaries,trends andforecasts

What is theaverageincome ofmutual fundbuyers, byregion, by year?

Queryand Reporting OLAP Data Mining

Insight & Prediction Information Analysis


10/61


Example Data Mining ApplicationsFinancial Services Combat attrition (churn)

Fraud detection Loan default (Basel II)

Identify selling opportunities

Database Marketing Buy product x More targeted & successful

campaigns Identify cross-sell & up-sell

opportunitiesTelecommunications

Identify customers likely to leaveTarget highest lifetime valuecustomers

Identify cross-sell opportunities

Insurance, Government Flag accounting anomalies

(Sarbanes-Oxley) Reduce cost of investigating

suspicious activity or false claims

Retail Loyalty programs Cross-sell Market-basket analysis Fraud detection

Life Sciences Find factors associated withhealthy or unhealthy patients Discover gene and protein targets Identify leads for new drugs


11/61


Oracle Data Mining 10gR2Oracle in-Database Mining Engine

Oracle Data Miner (GUI) Simplified, guided data mining

Spreadsheet Add-In for Predictive Analytics

1-click data mining from a spreadsheet PL/SQL API & Java (JDM) API Develop advanced analytical applications

Wide range of algorithms Anomaly detection Attribute importance Association rules

Clustering Classification & regression Nonnegative matrix factorization Structured & unstructured data (text mining) BLAST (life sciences similarity search algorithm)


12/61


13/61


14/61


15/61


Customer References"...Because Data Mining algorithms and the data arehoused together in the Oracle database, we don't have tomove huge data sets to external programs to run thealgorithms and learn something about our dataThe fact

that it cost about 75 percent less than the leadingcompetitor didn't hurt either "-- Tracy E. Thieret, Ph.D. Principal Scientist Xerox Innovation Group Imaging andSolutions Technology Center

Walter Reed Medical CenterUsing Oracle Data Mining, medical researchers arediscovering trends and patterns that will improve the healthcare for millions of people around the globe.--Dr. Carolyn Hamm, Director of Decision Support, Walter Reed Medical Center.Saving Lives with Oracle

IRS Detecting taxpayer noncompliance


16/61


Customer References

"Oracle Data Mining will allow us to pinpoint the mostimportant attributes of law school applicants thatcorrelate to successful legal careers. We will mine ourapplicant pool to seek our benefactors and trustees oftomorrow, therefore these strategic tools are critical toour long-term success. The security and scalability ofOracle's in-database mining, as well as its seamless

integration with our business intelligence platform weredeciding factors in selecting Oracle over analyticalalternatives."

-- Tom Delaney, CIO New York University School of Law


17/61


Data Warehousing

ETL

OLAP

Data Mining


Statistics

Oracle Data Mining 10gD E M O N S T R A T I O N


18/61


Oracle Data Mining Oracle Data Mining providessummary statistical informationprior to data mining


19/61


Oracle Data Mining

Oracle DataMiningsActivity

Guidessimplify &automatedata miningfor businessusers

Oracle Data Mining providesmodel performance andevaluation viewers


20/61


Oracle Data Mining

Additional model

evaluation viewers

Additional model

evaluation viewers

Apply model

viewers

E l #1


21/61


Example #1:Simple, Predictive SQL

Select customers who are more than 60% likely to

purchase a 6 month CD and display their maritalstatus

SELECT * from(SELECT A.CUST_ID, A.MARITAL_STATUS,PREDICTION_PROBABILITY(CD_BUYERS76485_DT, 1USING A.*) prob

FROM CBERGER.CD_BUYERS A)WHERE prob > 0.6;


22/61


Real-time Predictionwithrecords as (select

178255 ANNUAL_INCOME,0 CAPITAL_GAIN,83 SAVINGS_BALANCE,246 AVE_CHECKING_BALANCE,30 AGE,'Bach.' EDUCATION,'SelfENI' WORKCLASS,'Married' MARITAL_STATUS,'Sales' OCCUPATION,'Husband' RELATIONSHIP,'White' RACE,'Male' SEX,70 HOURS_PER_WEEK,'?' NATIVE_COUNTRY,98 PAYROLL_DEDUCTION from dual)

select s.prediction prediction, s.probability probabilityfrom (

select PREDICTION_SET( CD_BUYERS76485_DT , 1 USING *) psetfrom records) t, TABLE(t.pset) s;

On-the-fly, single recordapply with new data (e.g.

from call center)


23/61


Real-time Prediction Multiple Modelswith records as (select

178255 ANNUAL_INCOME,

0 CAPITAL_GAIN,83 SAVINGS_BALANCE,246 AVE_CHECKING_BALANCE,

30 AGE,'Bach.' EDUCATION,'SelfENI' WORKCLASS,'Married' MARITAL_STATUS,'Sales' OCCUPATION,'Husband' RELATIONSHIP,'White' RACE,'Male' SEX,70 HOURS_PER_WEEK,'?' NATIVE_COUNTRY,98 PAYROLL_DEDUCTION from dual)

select t.*from (

select 'CAR_MODEL' MODEL , s1.prediction prediction, s1.probability probability,s1.probability*25000 as expected_revenue from (

select PREDICTION_SET(NBMODEL_JDM, 1 USING *) psetfrom records ) t1, TABLE(t1.pset) s1

UNIONselect 'MOTOCYCLE_MODEL' MODEL , s2.prediction prediction, s2.probability probability,

s1.probability*2000 as expected_revenue from (select PREDICTION_SET(ABNMODEL_JDM, 1 USING *) psetfrom records ) t2, TABLE(t2.pset) s2

UNION

select 'TRICYCLE_MODEL' MODEL , s3.prediction prediction, s3.probability probability,s1.probability*50 as expected_revenue from (select PREDICTION_SET(TREEMODEL_JDM, 1 USING *) psetfrom records ) t3, TABLE(t3.pset) s3

UNIONselect 'BICYCLE_MODEL' MODEL , s4.prediction prediction, s4.probability probability,

s1.probability*200 as expected_revenue from (select PREDICTION_SET(SVMCMODEL_JDM, 1 USING *) psetfrom records ) t4, TABLE(t4.pset) s4

) torder by t.expected_revenue desc;

On-the-fly, single recordapply with multiple

models; sort by

expected revenues


24/61


Predictive Analytics: ExplainPL/SQL Package

BEGINDBMS_PREDICTIVE_ANALYTICS.EXPLAIN(

data_table_name => 'CD_BUYERS',

explain_column_name => 'CD_BUYER',result_table_name => 'explain_result37');

END;/SELECT * FROM explain_result37;


25/61


Predictive Analytics: PredictPL/SQL Package

SET serveroutput ON

DECLAREv_accuracy NUMBER(10,9);

BEGINDBMS_PREDICTIVE_ANALYTICS.PREDICT ( ACCURACY => v_accuracy,DATA_TABLE_NAME => 'CD_BUYERS',CASE_ID_COLUMN_NAME => 'CUST_ID',TARGET_COLUMN_NAME => 'CD_BUYER',RESULT_TABLE_NAME => 'predict_result24');

DBMS_OUTPUT.PUT_LINE('Accuracy = ' || v_accuracy);END;

/SELECT * FROM predict_result24 WHERE rownum


26/61


Example #2Launch & Evaluate a Marketing Campaign

select responder, cust_region, count(*) as cnt,sum(post_purch pre_purch) as tot_increase,avg(post_purch pre_purch) as avg_increase,stats_t_test_paired(pre_purch, post_purch) as

significancefrom (

select cust_name, prediction(campaign_model using *) as responder,

sum(case when purchase_date < 15-Apr-2005 then purchase_amt else 0 end) as pre_purch,

sum(case when purchase_date >= 15-Apr-2005 then purchase_amt else 0 end) as post_purch

from customers, sales, products@PRODDBwhere sales.cust_id = customers.cust_id

and purchase_date between 15-Jan-2005 and 14-Jul-2005

and sales.prod_id = products.prod_idand contains(prod_description, DVD) > 0

group by cust_id, prediction(campaign_model using *) )group by rollup responder, cust_region order by 4 desc;

1.Given a previouslybuilt responsemodel, predictwho will respond toa campaign,and why

2.find out howmuch eachcustomer spent 3months before andafter the campaign

3.how much for

just DVDs ?4. Is the success

statisticallysignificant?

O l D t Mi i


27/61


Oracle Data MiningAlgorithms & Example Applications

Attribute Importance Identify most influential attributes

for a target attribute Factors associated with high costs,

responding to an offer, etc.Classification and Prediction Predict customers most likely to:

Respond to a campaign or offer Incur the highest costs

Target your best customers Develop customer profiles

Regression Predict a numeric value

Predict a purchase amount or cost Predict the value of a home

A1 A2 A3 A4 A5 A6 A7

Income

Gender

Status Gender HH Size

>$50K 4

Age

Buy = 0 Buy = 1 Buy = 1 Buy = 0


28/61



Clustering Find naturally occurring groups

Market segmentation

Find disease subgroups Distinguish normal from non-normal behavior

Association Rules Find co-occurring items in a market basket

Suggest product combinations Design better item placement on shelves

Feature Extraction Reduce a large dataset into representative

new attributes Useful for clustering and text mining

F1 F2 F3 F4

Oracle Data Mining


29/61



Text Mining Combine data and text for better models

Add unstructured text e.g. physicians notes to

structured data e.g. age, weight, height, etc., topredict outcomes

Classify and cluster documents Combined with Oracle Text to develop

advanced text mining applications e.g. Medline

BLAST Sequence matching and alignment

Find genes and proteins thatare similar

ATGCAATGCCAGGATTTCCA

CTGCAA GGCCAGGA AG TTCCAATGC GT TGCCA C ATTTCCA

GGC.. TGCAATGCCAGGAT GA CCAATGCAATG TT AGGA CC TCCA

Oracle Data Mining 10g R2


30/61


Oracle Data Mining 10g R2Decision Trees

Problem: Find customerslikely to buy a new car andtheir profiles Decision Trees

Classification

Prediction Customerprofiling

Income

Gender

Status Gender HH Size

>$50K 4

Age

Buy = 0 Buy = 1 Buy = 1 Buy = 0

50K AND Gender=F AND Status >Single ), THEN P(Buy Car=1)Confidence= .77

Support = 250

Oracle Data Mining 10g R2


31/61


Oracle Data Mining 10g R2Anomaly Detection

Problem: Detectrare cases One-Class SVM Models

Fraud, noncompliance

Outlier detection Network intrusion detection Disease outbreaks Rare events, true novelty

X2X1

X2X1

Oracle Data Mining


32/61


Oracle Data MiningAlgorithm Summary 10gR2

Classification

Association Rules

Clustering

Attribute Importance

Problem Algorithm Applicability

Adaptive Bayes Network

Nave BayesPopular / Rules / transparency

Embedded app

Minimum DescriptionLength (MDL)

Attribute reductionIdentify useful dataReduce data noise

Hierarchical K-Means

Hierarchical O-Cluster

Product groupingText miningGene and protein analysis

AprioriMarket basket analysisLink analysis

Support Vector Machine Wide / narrow data

Support Vector Machine Wide / narrow dataRegression

Feature Extraction NMFText analysis

Feature reduction

Decision Tree

Rules / transparency


33/61


34/61


Integration with Oracle BI EE

Likelihood to buy


35/61

i i h O l i


36/61


Integration with Oracle Discoverer



37/61


Spreadsheet Add-In for Predictive Analytics

Enables Excelusers to mineOracle or Exceldata using oneclick Predict andExplain predictiveanalytics features

Users select a tableor view, or point todata in Excel, and

select a targetattribute


38/61


Oracle Data Mining & Oracle Text

Oracle Data Miningmines text to buildclassification and

clustering models Oracle Text(included in Oracle DatabaseStandard Edition)

preprocessesunstructured text

Handles large

volumes ofdocuments or text


39/61


Data Warehousing

ETL

OLAP

Data Mining


Statistics

Oracle Data Miner 10gR2Code Generation Release

Oracle Data Miner (gui)10 R2 S OTN R l


40/61


10gR2 Summer OTN Release

PL/SQL codegeneration for

Mining Activities



41/61





42/61





43/61



Oracle Data Miner (gui)10gR2 S mmer OTN Release


44/61




45/61


Data Warehousing

ETL

OLAP

Data Mining


Statistics

In-Database Analytics

Example

Example #1T M k i C i


46/61


Test a Marketing Campaign

Given a previously built response model

(classification), predict who will respond tothe campaign, and why

Example #1P di t R d


47/61


Predict Responders

select cust_name, prediction(campaign_model using *)

as responder, prediction_details(campaign_model using *)

as reasonfrom customers;

Example #1Combine with Relational Data


48/61


Combine with Relational Data

In addition to predicting responders, find

out how much each customer has spentfor a period of 3 months before and afterthe start of the campaign

Example #1Combine with Relational Data


49/61


Combine with Relational Data

select cust_name, prediction(campaign_model using *) as

responder,

sum(case when purchase_date < 15-Apr-2005 then purchase_amt else 0 end) as pre_purch,sum(case when purchase_date >= 15-Apr-2005

then purchase_amt else 0 end) as post_purch

from customers , saleswhere sales.cust_id = customers.cust_id

and purchase_date between 15-Jan-2005 and 14-Jul-2005

group by cust_id, prediction(campaign_model using *);


50/61

Example #1Multi-Domain Multi-DB data


51/61


Multi-Domain, Multi-DB data

select cust_name, prediction(campaign_model using *) as responder,

sum(case when purchase_date < 15-Apr-2005 then purchase_amt else 0 end) as pre_purch,

sum(case when purchase_date >= 15-Apr-2005 then purchase_amt else 0 end) as post_purch

from customers, sales , products@PRODDBwhere sales.cust_id = customers.cust_id

and purchase_date between 15-Jan-2005 and 14-Jul-2005and sales.prod_id = products.prod_idand contains(prod_description, DVD) > 0

group by cust_id, prediction(campaign_model using *);

Example #1Test Effectiveness / Significance


52/61


Test Effectiveness / Significance

In addition to predicting responders, find out howmuch each customer has spent on DVDs for a

period of 3 months before and after the start ofthe campaign, and Compare the success rate of predicted

responders and non-responders within differentregions and across the company

Is the success statistically significant?

Example #1Test Effectiveness / Significance


53/61


Test Effectiveness / Significance select responder, cust_region, count(*) as cnt,

sum(post_purch pre_purch) as tot_increase,avg(post_purch pre_purch) as avg_increase,stats_t_test_paired(pre_purch, post_purch) as

significance

from (select cust_name,

prediction(campaign_model using *) as responder,sum(case when purchase_date < 15-Apr-2005 then

purchase_amt else 0 end) as pre_purch,sum(case when purchase_date >= 15-Apr-2005 then

purchase_amt else 0 end) as post_purchfrom customers, sales, products@PRODDBwhere sales.cust_id = customers.cust_id

and purchase_date between 15-Jan-2005 and 14-Jul-2005and sales.prod_id = products.prod_idand contains(prod_description, DVD) > 0

group by cust_id, prediction(campaign_model using *) )group by rollup responder, cust_region order by 4 desc;

Analytics vs.


54/61


1. In-Database Analytics EngineBasic Statistics (Free)Data MiningText Mining

2. DevelopmentPlatform

Java (standard)

SQL (standard)J2EE (standard)

3. Costs (ODM: $20K cpu)Simplified environmentSingle serverSecurity

1. External Analytical EngineBasic StatisticsData MiningText Mining (separate: SAS EM for Text)Advanced Statistics

2. DevelopmentPlatform

SAS Code (proprietary)

3. Costs (SAS EM: $150K/5 users)Annual Renewal Fee

(~40% each year)

Data Warehousing

ETL

OLAP

Data Mining


Statistics


55/61

Advanced Analytics Partners


56/61


Advanced Analytics Partners

Benefits of Oracles Approach


57/61


In-Database Analytics Benefit Platform for AnalyticalApplications

Eliminates data movement andsecurity exposure

Fastest: Data Information

Wide range of data miningalgorithms & statisticalfunctions

Supports most analyticalproblems

Runs on multiple platforms Applications may be developedand deployed

Built on Oracle Technology Grid, RAC, integrated BI, SQL & PL/SQL available Leverage existing skills

Oracle 10Oracle 10 gg DBDBOracle Advanced Analytics


58/61


Know More! Leverage your data, discover new hidden

information and valuable insights, and makepredictions

Do More! Build applications that automate the extraction and dissemination

of data minings insights Move from End User Tool to Enterprise BI Application

Spend Less! Option to Oracle 10g Database Enterprise Edition Eliminates need for redundant data, new servers, new software,

and new support skills/resources

Data Warehousing

ETL

OLAP

Data Mining


Statistics

y

For More Information


59/61


Oracle Business Intelligence Solutions oracle.com/bi

Oracle Data Mining 10g oracle.com/technology/products/bi/odm/index.html


60/61

Q U E S T I O N SQ U E S T I O N S

A N S W E R SA N S W E R S


61/61

This presentation is for informational purposes only and may not be incorporated into a contract or agreement.

Documents

Pres Db Data Mining 10gr2 0806