Churn Modeling

Binary classification using embeddings of categorical features.

Dataset:

The dataset used for this experiment is provided in https://github.com/sharmaroshan/Churn-Modelling-Dataset Balance of Data: {1: 0.2037, 0: 0.7963}

Pipelines:

1. Package categorical features together and separate those from numerical features.
2. Package each categorical features separately and separate from numerical features.

Modeling:

Used both Tensorflow and Pytorch to achieve an 85% accuracy on test data. The aim was to compare the performance of these two packages on the same dataset.

It would be nice to compare with the performance of tree-based models.