Introduction to Machine Learning Models Using IBM SPSS Modeler (V18.2) SPVC [0E079G]

OVERVIEW

This course provides an introduction to supervised models, unsupervised models, and association models.

This is an application-oriented course and examples include predicting whether customers cancel their subscription, predicting property values, segment customers based on usage, and market basket analysis.

This course contains a PDF course guide, as well as a lab environment where students can work through demonstrations and exercises at their own pace.

OBJECTIVES

After this course participants should be able to:

• Describe the machine learning models
• Describbe Supervised models Decision trees :
• • CHAID
  • C&R Tree 
• Evaluate measures for supervised models
• Describe Supervised models:
• • Statistical models for continuous targets - Linear regression 
  • Statistical models for categorical targets - Logistic regression
• Associate models: Sequence detection
• Describe Supervised models:
• • Black box models - Neural networks 
  • Black box models - Ensemble models 
• Describe Unsupervised models:
• • K-Means and Kohonen
  • TwoStep and Anomaly detection  
• Describe Association models:
• • Apriori
• Prepare data for modeling

CONTENT

Introduction to machine learning models

• Taxonomy of machine learning models 
• Identify measurement levels 
• Taxonomy of supervised models 
• Build and apply models in IBM SPSS Modeler  

Supervised models: Decision trees - CHAID 

• CHAID basics for categorical targets 
• Include categorical and continuous predictors 
• CHAID basics for continuous targets 
• Treatment of missing values  

Supervised models: Decision trees - C&R Tree 

• C&R Tree basics for categorical targets 
•  Include categorical and continuous predictors 
• C&R Tree basics for continuous targets 
• Treatment of missing values  

Evaluation measures for supervised models 

• Evaluation measures for categorical targets 
• Evaluation measures for continuous targets  

Supervised models: Statistical models for continuous targets - Linear regression 

• Linear regression basics 
• Include categorical predictors 
• Treatment of missing values  

Supervised models: Statistical models for categorical targets - Logistic regression

• Logistic regression basics 
• Include categorical predictors 
• Treatment of missing values

Association models: Sequence detection 

• Sequence detection basics 
• Treatment of missing values

Supervised models: Black box models - Neural networks 

• Neural network basics 
• Include categorical and continuous predictors 
• Treatment of missing values  

Supervised models: Black box models - Ensemble models 

• Ensemble models basics 
• Improve accuracy and generalizability by boosting and bagging 
• Ensemble the best models  

Unsupervised models: K-Means and Kohonen 

• K-Means basics 
• Include categorical inputs in K-Means 
• Treatment of missing values in K-Means 
• Kohonen networks basics 
• Treatment of missing values in Kohonen  

Unsupervised models: TwoStep and Anomaly detection 

• TwoStep basics 
• TwoStep assumptions 
• Find the best segmentation model automatically 
• Anomaly detection basics 
• Treatment of missing values  

Association models: Apriori 

• Apriori basics 
• Evaluation measures 
• Treatment of missing values

Preparing data for modeling 

• Examine the quality of the data 
• Select important predictors 
• Balance the data