A primary goal of predictive modeling is to find a reliable and effective predictive relationship between an available set of features and outcome. Ineffective feature representations and the inclusion of irrelevant features are two key data characteristics that can prevent a model from demonstrating good performance. Feature engineering and selection: a practical approach for predictive models provides an extensive set of techniques for uncovering effective representations of the features for modeling the outcome and for finding an optimal subset of features to improve a model’s predictive performance.
Key features of this text
Provides start to finish guidance for the process of building and improving predictive models
Uses data which contain realistic challenges that practitioners regularly encounter
Presents feature engineering techniques that are both pragmatic and novel and that cover a wide variety of predictor types found in most data
Pinpoints common and subtle pitfalls of feature selection for the practitioner to consider
Demonstrates effective ways of implementing several greedy and global feature selection methods.
Supplies all data sets and analysis code on GitHub so that results can be reproduced.