Beruflich Dokumente
Kultur Dokumente
18-July-2016
Evaluation
Component
Weight (%)
Group Assignment
20
35
45
What else
Attendance
Grades
Exams (open book / closed book ?)
Assignments
CR
Group Assignments
8-10 groups
Few in-class assignments
MS-Excel, XLMiner, SPSS, R, etc.
Data Mining
Data mining is the process of
discovering meaningful new correlations, patterns
and trends
by sifting through large amounts of data stored in
repositories
using pattern recognition technologies as well as
statistical and mathematical techniques
(Gartner Group, as reported in Data Mining book)
Business Analytics
Analytics
Extensive use of data, statistical and quantitative
analysis
Explanatory and predictive models
Fact-based analytical decision making
Analytical competitor
Categories of Analytics
Descriptive analytics
Analytics for
CRM
customer segmentation and targeting
understand customer behavior and predict needs
customer acquisition and retention
Human Resources
Hiring and employee retention
Models to predict
a long-distance customers likelihood of switching to a
competitor
loyal versus disloyal
No Response
Variables
Continuous
Predictor
Variables
Linear Regression
K-nearest neighbors
Logistic Regression
Discriminant Analysis
K-nearest neighbors
Principal
components
Cluster analysis
Categorical
Predictor
Variables
Linear Regression
Regression trees
Classification trees
Logistic Regression
Association rules
Steps
Organizing datasets
Sampling, over-sampling
Preprocessing and data-cleaning
Categorical variables coding
Variable selection
Outliers, missing values
Normalizing, standardization
Partitions
Training
Validation
Test
New data
Capturing Data
Nominal
Ordinal
Interval
Ratio
Capturing Data
Categorical data
Nominal
Ordinal
Numerical data
Interval
Ratio
IV / DV
Independent variable
Dependent variable
Input variable
Predictor variable
Response variable
Attribute, Feature
Target variable
Topics
Simple methods
Association rules
kNN
RFM
Predictive Modeling
Credit Scoring/Rating
To assess the credit worthiness of a loan applicant
What are the chances that the borrower will default
a payment for mortgage, credit card, or other loan
Models need to identify characteristics that are
associated with credit worthiness
Fraud Detection
Credit cards: 10,000 payment transactions per
second
Statistical techniques
Calculating statistical parameters (averages,
performance metrics, probability distributions) to track
deviations (e.g. average length of call, average number
of calls per month)
Models and probability distributions of various
business activities
Clustering and classification to find patterns and
associations
Fraud Detection
Artificial intelligence
Expert systems
Pattern recognition to match given inputs
Machine learning techniques using sample
suspicious cases
Insurance
Analyze and predict customer attrition
Recruiting and monitoring agents
Lifetime value of customers and profitability
analysis
Cases of fraud
Risk models for loss estimation and portfolio risk
CRM
Customer acquisition, retention and
expansion
RFM (Recency-Frequency-Monetary value) cube
Customer Life-time Value (profitability analysis)
What-if scenarios
Predictive modeling
Personalization
Revenue Analytics
Optimal Pricing: Demand forecast, competitor rates,
and price sensitivity
Hotel rooms (airlines)
Rooms unsold, how many guests are expected, any special
events, how are competitors doing, is this high season or
low season, any regular customers
Season, day of the week, competition, pickup/denial (with
price changes), promotions and customer feedback
Supply Chain
Actionable insights to:
Understand inventory trends (optimize inventory
levels)
Track vendor performance to develop relationship
Quick response to market information
Analyze supply chain efficiency
Talent Analytics *
Human Capital
Key indicators for overall organizational health
Investment actions (recruitment and retention)
Workforce forecasts
Predict future headcount for each business unit, and guide
recruitment decisions
Retail
Customer Profiles
Identify best customers, profitability analysis
Targeting
Whom to target, communication channel and content;
response to direct marketing
Pricing
Demand forecasting, Price optimization and promotions
Sports
Sabermetrics
Objective knowledge about baseball using
statistics
SABR (Society for American Baseball Research)
To predict player performance based on
current form and related statistics
Useful in team selection and development.
Other
Text analytics
Linguistic, statistical and machine learning
techniques
Extract meaningful information from textual
sources like customer feedback, emails,
conversation transcripts, blogs, etc.
Web analytics
visits, page views, pages per visit
bounce rates and average time on site
No Response
Variables
Continuous
Predictor
Variables
Linear Regression
K-nearest neighbors
Logistic Regression
Discriminant Analysis
K-nearest neighbors
Principal
components
Cluster analysis
Categorical
Predictor
Variables
Linear Regression
Regression trees
Classification trees
Logistic Regression
Association rules
Analytical Capability
The Five Stages of Analytical Capability
Analytically Impaired
Localized Analytics
Analytical Aspirations
Analytical Companies
Analytical Competitors
* Davenport, Harris, Morison in Analytics At Work (Harvard Business Press, 2010)
Competing on Analytics
Widespread use of modeling and optimization
(recurring decisions are automated)
Enterprise-wide approach (information based
strategy)
Top management support (knowledge culture, right
processes, skills)
Privacy issues
Stalking Customers
Incorrect data
Data Security
Learned response from customers
Thank you