Innovate Analytics with Oracle Data Mining & Oracle R

Better intelligence, smarter decisions

Christian Screen, CapgeminiOracle Analytics Practice (formerly BI Consulting Group)

Tuesday December 11, 2012

Innovate Analytics with Oracle Data Mining & R

| Better intelligence, smarter decisions

Christian Screen

© 2012 Capgemini. All rights reserved.

o Solutions Engineer at Capgemini o15 Years in Technologyo Co-Author of Oracle BIEE 11g – A Hands-On Tutorialo Podcast & Blog at ArtofBi.como Oracle ACEo Oracle Deputy CTOo Oracle Hyperion Certified Consultanto BI Evangelist


Agendao What Data Mining?

o What is R?

o Oracle BI 11g + Oracle Data Mining!

o Use Cases

o Getting your Organization Started

o Oracle Endeca & Data Mining

o With Oracle BI 11g

o Oracle Predictive Analytics stack - Tie it all together

o Q & A



Getting the Message Today

An understanding of Data Mining and RHow Predictive Analytics is what your organization needs nowMotivation for starting a Predictive Analytics projectHow Oracle BI 11g works with Data Mining & R

What I want to leave you with today…



What is Data Mining?

“The process that attempts to discover

patterns and hidden knowledge in large data sets in order to

aid the decision making process.”

DiscoveryPattern

s Fraud Detection

Algorithms

Clustering

Classification

Association Analysis

Predictive Analytics

Regression

Summarization

Affinities

ProbabilitiesStatistical Confidence

Un

bia

se

dLarge Data Sets

Anomaly Detection



Data Mining Use Cases

Example # 1 – Market Basket Analysis

Determine combination of items purchased together generating highest or lowest margin

Predict new item sales when sold together with product sold at discount price

Calculate based on previous sales of individual products which store would benefit most from joined product sales

Example: computer + monitor, printer + paper

Example #2 – Relationship Patterns

Think Linked-In and Facebook Based on similar existing data points

locate other accounts and rank based on affinity

Predict how may friends/people may accept an offer based on previous behavior and the behavior of similar users



Oracle Data Mining

Oracle’s DM Solution for Solving Business Problems

Oracle Data Mining is installed automatically when you install Oracle Database Enterprise Edition.

GUI is included with SQL DeveloperA mature DM Platform



What is R?

o A Open Source programming language and software environment

for statistical computing and graphics.

o Standard developed and maintained by the R Foundation

o Part of the Free Software Foundation GNU project

o Watch Movie: Revolution OS (2002)

o Client or Client Server

o Many images available on Amazon EC2

o R-Studio seems to be a leading Open Source IDE for R



What is R?

What Else?o Really good at statistical and graphical techniques, including linear

and nonlinear modeling, classical statistical tests, time-series

analysis, classification, clustering, and others

o Text Mining and Lexical scoping

o Extends via user created Libraries/Packages

o Automated report and document generation

o Think associations, algorithms, statistics and

surprisingly good graphics



Data Mining vs. R

R

Does not require a database Complex algorithms can require large

amounts of memory and processor Visual output built in or by libraries User-built libraries make for expansive

capabilities and logic to be developed Skill-up required for R Script language Data does not need to be clean

though it helps Uses connectors for data sources

Data Mining

Develop Statistical Models Resides in the Database No visual output SQL Developer plug-in available Leverage some existing DB skills Likes cleansed data



Oracle BI & Data Mining / R

Start today! Think Competitive Advantage.Data Scientists are heroesExperimentation is keyLeverage existing investments Its part of Business IntelligenceUsed by a few, developed by fewerGet Endeca Information DiscoveryGet R & R-StudioExplore Predictive Analytics

Getting the Organization Started is Simple



Oracle BI & Data Mining / R

Traditional BI is necessary but can become stagnantPredictive Analytics complements BIExisting BI investments distribute Predictive AnalyticsAll areas of an organization can benefit

Bringing it All Under Business Intelligence makes sense…



Oracle Predictive Analytics Now

BI, Discovery, Data Mining, & R

Predictive Analytics

Includes traditional BI concepts

Expands the BI TeamExpands knowledge of

informationTactically targets specific

scenarios of the businessProvides non-linear

perspective on the business

Tells us how we’re doing today

Distributes informationTracks Performance MetricsMakes sense of ERP data

Traditional BI / Analytics



Oracle Predictive Analytics

Endeca is mainly a three part tool (ETL, Server, Client)• Data Mining typically depends on a cleansed source, i.e. DW

Endeca includes powerful features, visualizations, and guided search navigation, perfect for self-service discovery• Data Mining is database driven so no inherent graphics• R has visuals and client tools for development but is not distributed

Endeca projects perform well in an iterative design-develop-feedback loop• Data Mining and R are built from models and algorithms are typically consumed

by other systems

Endeca Information Discovery (EID) vs. Oracle Data Mining / R



Oracle Predictive Analytics Stack

Oracle Advanced Analytics in Oracle 11g R2 RDBMS• Data Mining & Enterprise R

o Aims to eliminate need for SAS, SPSS, Matlab, etc.

Enterprise-R• Oracle’s flavor of R

Exadata MachineExalytics (as consumer)

• Endeca• OBIEE• Essbase

Engineered to work together…

Start Planning Now for Running your business in

the future of Analytics.



Upcoming Endeca Webinar

Predictive Advertising Analytics in MediaScott Schlesinger Paper on Predictive Analytics in the Movies

(http://www.digitalmarketingsuite.com/url/60534)

Endeca Information Discovery in the Wild



Data Mining with Oracle BI 11gOracle R Enterprise in Oracle BI Dashboards



Data Mining with Oracle BI 11g



Oracle R with Oracle BI 11gORE in Oracle BI 11g dashboard – Flight Delays



Data Mining & R with Oracle BI 11g

Embedded R Script using Parameterization from Oracle BI select * from table(rqTableEval( cursor(select ARRDELAY,DISTANCE,DEPDELAY,YEAR,MONTH,DAYOFMONTH,DEPTIME,ARRTIME,UNIQUECARRIER,FLIGHTNUM,ORIGIN,DEST, ORIGIN||'-'||DEST ROUTE from ontime_s where year >= valueof(NQ_SESSION.OR_ARG1) and year <= valueof(NQ_SESSION.OR_ARG2) and DEPDELAY is not NULL), cursor(select 1 max1, 1 pos1, 'mod' name1, to_number(null) max2, to_number(null) pos2, to_char(null) name2, total, chunk, value from ontime_lm), 'select ARRDELAY, DISTANCE, DEPDELAY,YEAR,MONTH,DAYOFMONTH,DEPTIME,ARRTIME,UNIQUECARRIER,FLIGHTNUM,ORIGIN,DEST, ORIGIN||''-''||DEST ROUTE, 1 PRED from ontime_s', 'PredictDelays-score'))order by 1, 2, 3


Better intelligence, smarter decisions

Questions?


www.capgemini.com/bim

The information contained in this presentation is proprietary. ©2012 Capgemini. All rights reserved

Contact Christian [email protected]

Access Oracle BI Training at Capgemini’s BICG Universitylink

mailto:[email protected]

http://www.biconsultinggroup.com/bicg-university/

Innovate Analytics with Oracle Data Mining & Oracle R

Technology

r r data mining

oracle bi data mining

oracle data miningoracles

o oracle bi

understanding of data

smarter decisions

data scientists

data mining use casesexample