dheerajnbhat / animated_octo_guide

Petfinder Classification task

This repository contains code to train XGBoost model using Petfinder dataset.

Requirements

• Python3.10 (tested on python3.10)
• conda (optional)

Installation

Install using conda

If you have conda, go to the root of the repo and use the following command to create a conda environment:

conda env create -n myenv -f environment.yml
conda activate myenv

Install using python virtual environment

If you are using python virtual environment, go to root of the repo after activating the environment, and install the dependencies using the following command:

pip install -r requirements.txt

Directory structure

├── artifacts # contains model and charts

├── tests # contain pytests for data, training and predictions

├── test.sh # script to run tests

├── app.py # streamlit demo

├── environment.yml # conda environment file

├── requirements.txt # python dependencies file

├── evaluation.py # python script to evaluate model

├── evaluation.sh # bash script to run evaluation.py

├── train.py # python script to train the model

├── train.sh # bash script to run train.py

├── run.sh # script to run training, evaluation and tests at once

└── ...

Running the script

To run the entire script, go to the root of the repo and use the following command:

./run.sh

To run individual scripts, use the following command:

./train.sh
./evaluation.sh
./test.sh

train.sh

• Reads the input from "gs://cloud-samples-data/ai-platform-unified/datasets/tabular/petfinder-tabular-classification.csv" and loads into Pandas DataFrame
• Split the dataset into 3 splits: train, validation and test with ratios of 60 (train) / 20 (validation) / 20 (test)
• target column: Adopted, input features: rest all columns
• Train ML model using XGBoost to predict whether a pet will be adopted or not. Uses validation data to perform early stopping
• Log the training process to train.log with accuracy, f1, recall and precision score evaluated on test split of dataset

evaluation.sh

• Reads the input from "gs://cloud-samples-data/ai-platform-unified/datasets/tabular/petfinder-tabular-classification.csv" and loads into Pandas DataFrame
• Uses the model you trained in the previous step to score all the rows in the CSV
• Save the output into output/results.csv (output/* is git ignored)

test.sh

• running pytests

Demo

To run the demo, use the following command:

streamlit run app.py

Next steps

• Using DVC for storing artifacts
• Improving the model accuracy