Wednesday 4:50 PM–5:25 PM in Main Room

Feature Engineering, Feature Selection and Class Architecture Approaches to Build Optimal Models

Ben Fowler

Audience level:


Model success can be predicated by feature engineering, feature selection and class architecture. Feature engineering involves wrangling of data to enhance predictive power. Feature selection allows the machine to identify the features that are most important in the model; which allows the model to generalize to unseen data. Optimizing class architecture can improve the misclassification rate.


In this talk, the following concepts will be reviewed:

- Feature Engineering Approaches With:

1) Categorical Features

2) Numerical Features

3) Dates

4) Time Series

- Feature Selection Approaches:

1) Forward Feature Selection

2) Backward Feature Selection

- Class Architecture Approaches:

1) Handling Class Imbalance

2) How Many Classes

3) Class Separation Thresholds

Case Study Example: Fantasy Football

Subscribe to Receive PyData Updates



Get Now