View the deployed application here: https://xgb-digit-classifier.herokuapp.com/
This repo demonstrates how to build a classifier using
- mnist dataset: https://keras.io/api/datasets/mnist/
- ensemble classifiers (random forest, xgboost)
Important files:
initial-model-training.ipynb
: preprocess the dataset, train the modelsoptimization
: this folder contains alternative versions of the trained models, using gridsearchrf_grid_model.tar.gz
: this is a compressed file containing thepkl
file because the original was too large for github.
I was inspired by these other apps: