Making Sense of Data: A Practical Guide to Exploratory Data Analysis and Data Mining

7.7 SUMMARY

Types of models:

• Classification: Models where the response variable is categorical. These are assessed using: concordance, error rate, specificity, and sensitivity analysis.

  Table 7.20. Summary of predictive modeling approaches in this chapter

  

• Regression: Models where the response variable is continuous. These are assessed using r² and residual analysis.

Building a prediction model involves the following steps:

1. Select methods based on problem
2. Separate out training and test sets
3. Optimize the models
4. Assess models generated

Applying a prediction model follows these steps:

1. Evaluate whether an observation can be used with the model
2. Present observations to model
3. Combine results from multiple models (if appropriate)
4. Understand confidence and/or explain how results were computed

Table 7.20 summarizes the different methods described in this chapter.