Introduction to Predictive Analytics

Rating:
1 vote, average: 5.00 out of 51 vote, average: 5.00 out of 51 vote, average: 5.00 out of 51 vote, average: 5.00 out of 51 vote, average: 5.00 out of 5
Loading...
Please Log in or register to rate

Introduction to Predictive Analytics

BD-107

Traditional data analysis is about describing the data we have. Some more advanced models are used to explain why we get what we see. However, with advances in technology and statistical theory, most companies now employ (or can employ) a new statistical paradigm: predictive analytics: instead of finding characteristics of customers, predictive models can predict which one is likely to leave, or which lead is most likely to convert.

This change in paradigm benefits decision makers and managers as it provides more precise insights that lead to focused and valuable actions. But it also requires new statistical capabilities. It turns out that the best explanatory models are not always the best predictive models, and analysts now need to develop, evaluate and interpret their models differently.

In this workshop we will get an overview of the world of predictive models and analytics, enter into this new “mindset”, and learn the basic considerations and evaluation techniques. In a “hands on” manner we will learn to classify, estimate, cluster and predict outcomes in real world settings, using the R statistical environment.

The workshop is modular and built as a mix of interactive demonstrations of key topics, in-class exercises based on real business settings and “open audience consultations” in which participants bring their own data and receive advice on how to accomplish their goal. The typical workshop takes 3-4 full days, and specific topics can be tailored to the needs and background of participants. Some background in using R is required.

Audience

Target Audience:
The workshop is aimed at business analysts, statisticians and other professionals working with data, who need (or want) to use predictive models in their day-to-day analysis tasks.

Prerequisites:
Previous experience in analyzing data is required, as well as some knowledge of R.

Course Topics

Module 1: Mapping the Terrain

  • Descriptive, Explanatory and Predictive modeling
  • Example: why you need “predictive” regression
  • The kind of problems predictive models deal with
    –classification
    –estimation
    –clustering
    –rule and pattern extraction
    –time series prediction and time-to-event predictions
  • Supervised vs. unsupervised learning
  • Feature creation and feature selection

Module 2: The basic “mindset” and Evaluation Techniques

  • Cross validation and bootstrapping techniques
  • Estimation problems using regression
  • Regularization of models and its role (bias-variance trade-off)
  • Tree-based estimation

Module 3: Classification

  • Logistic regression
  • Decision trees
  • Random forest
  • Naïve-bayes

Module 4: Clustering and Pattern Analysis

  • k-means
  • Hierarchical clustering
  • Association rules and market basket analysis

Module 5: Feature Selection and Feature Extraction and other Value Enhancing Techniques

  • Principal component analysis
  • Practical consideration in creating and selecting features
  • The concept of boosting

Module 6: Common Business Applications

  • Customer analytics (customer churn, etc.)
  • Recommender systems
  • People analytics

**Additional Potentially relevant Modules

  • Time series models
  • Survival analysis and time-to-event predictions
  • Social network analysis
© Copyright - Skilit - Site by Dweb