Loading...

BUILDING STRATEGIES USING PREDICTIVE ANALYTICS

Course Details
Reporting, Dashboarding, Query and Analysis, Predictive Analytics
• What is it, how are they different
• What needs to be used where
Probability, Distributions and Hypothesis testing
• Probability, odds and score – how are they related
• P value and Significance for given distributions
Predictive Modeling Applications
• Targeting, propensity scoring, credit scoring, cross sell, up sell, churn management, collections
Business Problem Definition for Predictive Analytics
• Most underestimated and most crucial
• Don’t jump to solving it
• Person asking the problem does not know his problem
Sample Size
• Determining the impact of Sample size
• Sample size and confidence level account for statistical variance
Data Preparation for Predictive Analytics
• Merging data and bringing to the desired grain for the business problem
• Profiling, missing values, outliers – how to treat them
Predictive Analytics Techniques and their comparison
• Decision Trees
• Clustering
• Logistic Regression
• Scorecards
• Neural Networks (only overview-can go in depth if time left)
Tracking Performance and Tweaking Strategies
• External environment changes from Historical
• Marketing strategy and mix changes
• Predictive Analytics cannot avoid this but has to deal with it
Making Presentations and Gaining Visibility
• without presentation has no value
Significance of Predictive Analytics to Business
• How does Predictive modeling deliver high impact
• Why reporting/dashboarding does not work here
Course Duration
3 days – 9:30am to 4:30 pm

Detailed Contents.

Day 1

Reporting, Dashboarding, Query and Analysis, Predictive Analytics
Data Presentation section which outlines what method needs to be used for what purpose. What is it, how are they different – historical view vs future prediction; What needs to be used where – managing and accounting deviations from expected vs impacting profitability and risk.
Probability, Distributions and Hypothesis testing
Probability, odds and score – how are they related; type I and type II errors; P value and Significance for given distributions
Predictive Modeling Applications
Targeting, propensity scoring, credit scoring, cross sell, up sell, churn management, collections
Business Problem Definition for Predictive Analytics
Most underestimated and most crucial - Don’t jump to solving it; Person asking the problem does not know his problem – don’t allow any chance for communication gap, restate the problem in different ways; Ensure that is the root problem being solved and nobody has put layers of problems.
Sample Size
Determining the impact of Sample size; Sample size and confidence level account for statistical variance
Data Preparation for Predictive Analytics
Merging data and bringing to the desired grain for the business problem; Profiling, missing values, outliers, junk values, truncated data, business rules validation – how to treat them

Day 2

Predictive Analytics Techniques and their comparison
Defining timeframes- Decisioning, performance and readout; Decision Trees-Understanding information value/entropy; Applications- segmentation, identification of important variables, dealing with multiple objectives, multiple KPIs, marketing list generation, variable reduction; Logistic Regression and Scorecarding – Credit Scoring, Behavioral scorecards overview; Defining target variable; Choosing methods- forward, backward, stepwise; Dealing with collinearity- variance inflation factors; Scoring – base odds, pdo, confusion matrix, setting score cutoffs, when to reconstruct the score and when to change the strategy cutoffs, score matrix using multiple scores in strategy; Neural Networks for adapting models like fraud model; Methods that work well with smaller sample sizes, changing market dynamics, missing data, outliers.

Day 3

Tracking Performance and Tweaking Strategies
Loss Projection – roll rate method; Yield projection; unit risk and amount risk, risk based pricing – risk due to default, risk due to attrition, adverse selection; External environment changes from historical; Marketing strategy and mix changes; Predictive Analytics cannot avoid this but has to deal with it
Making Presentations and Gaining Visibility
Strategy without presentation has no value as it cannot be implemented; Supporting material; getting buyins before key meetings and working the room. Brevity is the essence - keep it short - 3 slide method;
Significance of Predictive Analytics to Business
How does Predictive modeling deliver high impact; Why reporting/dashboarding does not work here
Discussions on specific topics
Although we would be clearing doubts during the session, there would be about half a day kept at the end for clearing any topics that need to be dealt at depth.
FIRST NAME
LAST NAME
EMAIL
PHONE (with Country Code)
CITY
COUNTRY
Scroll Up
 Previous  All works Next