ML Example Project

This project was made as a good production ready example for a classical machine learning project. We tried to incorporate as many good ml habits and guidelines from the book Approaching (Almost) Any Machine Learning Problem by Abhishek Thakur and others.

The goal is to classify potable water using the data provided in here: https://www.kaggle.com/artimule/drinking-water-probability

Stacking

notebook

The best model was trained using the Stacking Generalization method where we first train a couple of weak learners, then use the predictions from these models to train a Meta-Model able to reduce the generalization error of the different weak learners and provide more robust predictions.

Interpretability

notebook

We used the SHAP (SHapley Additive exPlanations) method to explain the model's predictions. You can find a great explanation for this method in the interpretable ML book.

HuggingFace Space - Gradio

A demo app was made using Gradio, and published to a HuggingFace space, to enable users to play around with parameters and test the consistency of the model.

CI/CD

We also set up a Github workflow to automatically deploy the best model to the HuggingFace Space.

Carbon Emissions

These results are obtained using codecarbon. The carbon emission is estimated from performing GridSearch to select the best models. CodeCarbon: https://github.com/mlco2/codecarbon

timestamp	project_name	run_id	duration	emissions	emissions_rate	cpu_power	gpu_power	ram_power	cpu_energy	gpu_energy	ram_energy	energy_consumed	country_name	country_iso_code	region	cloud_provider	cloud_region	os	python_version	cpu_count	cpu_model	gpu_count	gpu_model	longitude	latitude	ram_total_size	tracking_mode	on_cloud
2021-11-22T22:53:21	codecarbon	c3d844aa-055f-46ff-922f-75d7c7ec1b9d	0.2960696220397949	1.1547931538621576e-08	3.9004108084649806e-05	22.5	0.0	3.100280832	2.5573372840881347e-08	0.0	3.523761672363281e-09	2.9097134513244628e-08	United States	USA				Linux-5.10.60.1-microsoft-standard-WSL2-x86_64-with-glibc2.31	3.9.5	8	Intel(R) Core(TM) i5-8300H CPU @ 2.30GHz	1	1 x NVIDIA GeForce GTX 1050	-97.822	37.751	8.267415552	machine	N
2021-11-22T22:57:06	codecarbon	9ef3dff7-c7b5-4562-adee-dff65eaedd95	128.58673238754272	0.0003395318887920674	0.0026404892828971265	22.5	0.0	3.1002808319999997	0.0007519074350595474	0.0	0.00010360552037126217	0.0008555129554308096	United States	USA				Linux-5.10.60.1-microsoft-standard-WSL2-x86_64-with-glibc2.31	3.9.5	8	Intel(R) Core(TM) i5-8300H CPU @ 2.30GHz	1	1 x NVIDIA GeForce GTX 1050	-97.822	37.751	8.267415552	machine	N

Usage

Initialization

Before running the models, you should put your data file (drinking_water_potability.csv) in the input/raw folder.

Option 1

To train and evaluate a model

./run.sh extratrees

Option 2

• To train a model:

python3 src/train.py --fold 0 --model extratrees

• To predict on new instances / evaluate a model:

python src/inference.py \
--model extratrees \
--data input/drinking_water_potability.csv