ML-project

Binary classification

Brief description

In this project, we aim to apply different machine learning models onto two different datasets for binary classification. However, our scripts can be applied to different datasets.

Algorithms

Preprocessing

• Loads datasets
• Cleans data
• Splits datasets into training and test datasets
• Additionnaly to cleaned data, outputs datasets on which we operated either tsne or PCA

Classification

We implement different models using Sklearn. The training is based on a GridSearch model that helps find the optimum parameters for a maximum accuracy across multiple validation sets using cross-validation. It contains the following models:

• SVM
• KNN
• Log regression
• Stochastic gradient descent

Other classification models

• Neural networks
• Random forests
• Decision Trees

All these models are trained using the gridsearch method.

Tests and results

In the program main.py, we output the best accuracies reached by each model (fitted using gridsearch) and the corresponding optimum parameters.

Authors

• Imane Momayiz
• Guillaume Sallé
• Romain Namyst
• Thomas Kronland
• Karel Kedemos