You are on page 1of 15

Analytics : Understanding Patterns

Tuesday 10 July 2012

The Universal Language of Measures

Tuesday 10 July 2012

Time

Proportions

Size

Financials

Productivity

Loyalty

The Universal Language of Cause & Effect

Process & Scale

Habits & Health

Technology & Efficiency

Consumer Understanding &


Pricing

Risk & Return

Action & Outcome

Tuesday 10 July 2012

Possibilities of no pattern unlikely ........


Cause

Effect = fn ( Data , Math , Common Sense)


Effect

Analytics is finding the relationship/ path of Cause to Effect

Tuesday 10 July 2012

Sources of Data

Tuesday 10 July 2012

Surveys

Transaction Systems

Free Text

Digital Images

Sensors

Voice

GPS

..... Upto the Imagination

Fundamental Concepts

Exponential Increase in
Computing Power

Explosion of Digitized Data

Democratization of
Multivariate Analytics ( NDimensional Plane )

Open Source Data Mining &


Statistical Software

Tuesday 10 July 2012

Tools For Data Mining & Predictive Modeling

Tuesday 10 July 2012

Universal Applications
Regression - Deriving Drivers

Cluster - Classifying & Grouping

Direct Marketing

Marketing

Scoring Applications

Customer Service

Forecasting

HR

Identifying critical influencing drivers

Across all functions....

Tuesday 10 July 2012

Evolution of Analytics - The Answers


Thought Analytics - You are how you think

2010

Sentiment Analytics - You are what you feel

2008

Social Media Analytics - You are the company you keep

2005

Transaction Data Analytics -You buy so you are

Pre 80s

Survey Analytics - Can I ask you?

Tuesday 10 July 2012

Evolution of Analytics - The Data & Techniques


Sensors / Artificial Intelligence

2010

Text /Voice/Imaging / Artificial Intelligence

2008

Web Logs / Text Mining/Multivariate

2005

Transaction Databases /Multivariate

Pre 80s

Questionnaire / Cross Tabs /Univariate /Bivariate

Tuesday 10 July 2012

Executing Analytics Projects


CRoss Industry Standard Process for Data Mining (CRISP-DM) for developing and deploying analytics
solutions
Problem
Objectives
Determine
Problem
objectives
Assess
situation
Determine
data mining
goals

Data
Study

Collect initial
data

Data
Preparation
Select data
Clean data

Describe data
Construct data
Explore data
Verify data
quality

Integrate data
Format data

Analysis &
Modeling

Reporting &
Evaluation
Deployment

Select analysis /
modeling
technique

Evaluate results

Plan deployment

Review process

Generate test
design

Determine next
steps

Plan monitoring
and maintenance

Build model

Produce final
report
Review project

Assess model

Produce
project plan
Domain expert
finalizes
objectives with
client

Tuesday 10 July 2012

Analysts use data


mining software to
integrate and
understand
relevant data

Complex data
cleansing
algorithms used to
collate all relevant
data into an
analytical data
mart.

Statisticians select
techniques) based on
hypothesis. Business
consultants and
analysts collaborate to
unearth key drivers and
forecast key business
indicators.

The solutions are


evaluated and
validated by the
business users and
practice head.

The solutions are


integrated with the
relevant business
processes.

Career Options
Geo Independent

Offshoring

Internal Client

Captives
BFSI/ Retail Captives

3rd Party ITES

Core
Analytics Division of Leading
Companies

Boutique

External Client
BI / Analytics Verticals of most ITES
firms

Products
Tuesday 10 July 2012

Small Companies Focused on Niche


Vertical & Function

Product Companies Like SAS/IBM- SPSS/ STATISTICA etc

Techniques of Data Mining - 1


Technique

Category

Description

Summarizing data

Data Understanding

Frequency counts of categorical


variables . Central Tendency Measures for
Numeric

Standardizing data

Data cleansing / Normalization

Format standardization , missing value


treatments
Integrating multiple databases to create
single database (datamart buildup )

Merging / Appending

Data Preparation

Variable Creation / Integration

Data Preparation

Creating Variables which the users


understand and derive meaning

Cross Tabulation

Reporting

High level reporting of 2*2 or more


variables

Cubes

Reporting

Multi level and real time drill downs of all


relevant variables

Macros

Automation

Automatic generations of all standard


reports / cubes.

Tuesday 10 July 2012

Techniques of Data Mining - 2


Technique

Category

Description

Measures of Central
Tendency

Data Understanding

Enables identifying the outliers and the central values

Hypothesis Testing /
Correlations

Analysis

Identification of whether basic assumptions related to


the data are valid or not . Used for simple analysis

Regressions/ Factor Analysis /


Predictive Modeling
ARIMA

Identifying the factors on which the key situation at


hand is dependent on. Forecasting Key Indicators

Clustering Models

Grouping / Segmentation

Bucketing records into mutually homogenous &


collectively heterogenous groups

Text Algorithms

Grouping

Preparing unstructured data to be in a form for


advanced statistical modeling

Artificial Intelligence/Neural
Networks

Inference and Judgement


Analytics

Building automated engines which analyze information


in a human simulated manner

Decision Trees/Chaid /SEM

Grouping / Segmentation

Root Cause Analysis , Path / Dependency Analysis

Tuesday 10 July 2012

Thank You

Tuesday 10 July 2012

You might also like