Using 8 years of data from the City of Austin's Animal Center, we predict the outcomes of cats held at the Center. A multinomial logistic regression gradient descent model predicts the 6 possible outcomes with about 50% accuracy. Notably, we code the model "by hand", using only matrix multiplication in numpy. The model can be improved further with more feature engineering and refined oversampling.
We describe our complete approach and results in a summary report and document our process in a notebook.