This project involves building a neural network model to predict employee attrition (whether an employee leaves the company) based on a variety of factors such as age, years worked, and monthly income.
The publicly available IBM employee attrition dataset is used. A 1-hidden-layer neural network will be built with a softmax one-hot output.
Exploratory data analysis is performed in IBM Attrition EDA.ipynb and the actual model is built up in network.py.
Running with tuned hyperparameters yields a peak accuracy of 83% before dropping off due to overfitting.