Scrambling for features!
To build a great machine learning model, it is very crucial to determine which features are the most important. Even after the data is cleaned and preprocessed, most algorithms perform better when fed with enginnered features that are specifically made to enhance the predictive properties of the dataset. A useful rule of thumb is that a fearture must have a relationship to the target in other to learn properly. Thus, feature engineering helps to achieve this by making your data better suited to the problem at hand.
In a nut shell
Our implemented methods will enable you to
- determine which features are the most important using mutual information
- invent new features that are suited to the problem at hand
- encode high-cardinality categoricals with a target encoding
- create segmentation features with k-means clustering
- decompose a dataset’s variation into features with principal component analysis
These procedures lead to an improved model’s predictive performance, reduced computational or data needs and improved interpretability of the results.
