preston4896/ecs171_pet_adoption

ECS 171 - Predicting Pet Adoption Speed

Group project for F19 ECS171 by: Amitpal Gill, April Vang, Brian Cirieco, Gan Qiu, Guanzi Yao, Pavlos Maltsev, Preston Ong, Stephanie Olivera, Xuan Deng, Zoe Kanavas

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. The src folder contains all source files, while data_management.py contains data analysis. The training/folder can be ignored (it was used for dev purposes).

scr/main.py and scr/main.ipynb are scripts that will run the whole code pipeline for you and report useful information as well as output all graphs. Each file can be run individually, or the main script will run everything in the correct order.

Prerequisites

In order to run the source files, please install:

• pip: astetik, numpy, pandas, scipy, tensorflow, keras, sklearn, talos, lightgbm
• Homebrew: libomp

Running

src/ files are meant to be run in the following order:

1. get_data_final.py
2. pca.py
3. HyperOptimization.py
4. test_ANN_model.py
5. pr_curve.py

• get_data_final.py will remove outliers, normalize and split data into training testing sets.
• pca.py is mainly use for visualization purposes
• HyperOptimization.py runs a grid search on the ANN ro determine best params. It is recommended not to run this file since it takes a long time to complete. The optimal hyper-parameters have been saved and are already used by the ANN (rerunning this script will override the current configuration).
• test_ANN_model.py runs the ANN with best params and RFE results, training/testing error & accuracy, and useful graphs.
• pr_curve.py will output ROC and AUC.

Again, src/main.py or scr/main.ipynb will initialize the correct pipeline and output all the information.