The repository contains the code for Machine Learning course 2020 (CS-433) project 1 (Higgs Boson challenge: https://www.aicrowd.com/challenges/epfl-machine-learning-higgs/leaderboards) at EPFL. More information about this challenge can be found in the folder documents.
The project is accomplished by team INteam with members:
- Riccardo Cadei: @RiccardoCadei
- Raphael Attias: @raphaelattias
- Shasha Jiang: @dust629
With a Test Accuracy of 0.841 we got the 7-th place out of 277 teams.
The data train.csv and test.csv should be found in https://github.com/epfml/ML_course/tree/master/projects/project1/data, to run the code please download and place them in the data folder
The project has been developed and test with python3.6.
The required library for running the models and training is numpy.
The library for visualization is matplotlib.
Results to predict the test datasets are generated by running:
python3 run.py.
And the final results are saved in: /data/finalsubmission.csv.
implementations.py: the implementation of 6 methods to train the model.
run.py: the results after using the selected model to predict test data.
exploration.py: understanding the features of data with visualization.
process_data.py: preprocessing data for model training and prediction.
crossvalidalidation.py: using cross-validation to test the accuracy of different models.
select_parameter.py: searching for the appropriate parameters(lambda, degree etc.) for models.
main.ipynb: tuning the best parameters for ridge regression and predicting the accuracy of all the methods through cross validation.
plots.ipynb: data analysis and visualizion of the accuracy and error with different choices of parameters.
documents/report.pdf: a 2-pages report of the complete solution.