AI Huasca Team: Henry Schultz 342151, Louis Tschanz 315774, Majandra Garcia 347470 [email protected] [email protected] [email protected]

Machine Learning Project

This project implements a comprehensive machine learning pipeline to preprocess data, train models, and evaluate their performance through cross-validation and hyperparameter tuning. It is structured with modular components for flexibility and reuse, making it suitable for various machine learning tasks.

Project Structure

The project is organized as follows:

• helpers_perso/
  Contains custom helper functions that facilitate specific tasks in the pipeline.

  • helpers_implementations.py: Provides utility functions that support core model implementations, including commonly used mathematical and data manipulation functions.
  • helpers_nan_imputation.py: Focuses on handling missing values, offering methods to impute NaN values based on various strategies such as mean, median, or mode.

• preprocessing/
  This folder houses modules for data preprocessing.

  • class_balancing.py: Contains the method to handle the imbalance of the dataset.
  • nan_imputation.py: Provides functions to impute missing values, allowing the user to specify imputation strategies to handle incomplete data effectively.
  • binary_encoding.py: Handles categorical data encoding, transforming categorical variables into a format that can be provided to ML algorithms.
  • remove_highly_correlated_features.py: Identifies and removes features that are highly correlated, reducing redundancy and potential model overfitting. (not used in the final pipeline)
  • standardization.py: Standardizes quantitative features.
  • preprocessing.ipynb: A Jupyter Notebook used to develop and tune the preprocessing steps, providing an interactive environment for testing and visualization.

• implementations.py
  This file contains the main implementations of the machine learning algorithms used in this project.

• run.py
  The main script for executing the project pipeline. It integrates preprocessing, model training, classification and submission file writing.

• crossvalidation.py
  Implements cross-validation functions for model assessment.

• Cross_Validation_and_Tuning.ipynb
  A Jupyter Notebook dedicated to exploring various hyperparameter tuning techniques and cross-validation results. It provides an interactive environment to visualize and analyze model performance across different hyperparameter configurations.

• predict_labels.py
  Contains functions specifically for generating predictions on new data, translating model outputs into labels that can be interpreted in the context of the application.

• prediction_score.py
  Evaluates model predictions by calculating metrics such as accuracy, precision, recall, or F1 score. This script helps assess model effectiveness and provides insights into areas for improvement.

Getting Started

Prerequisites

Ensure the following packages are installed:

• numpy
• matplotlib

Install any missing dependencies via pip install.

Running the Project

1. Data Preprocessing: Use preprocessing.ipynb to experiment on preprocessing of your dataset or run individual scripts within the preprocessing/ folder.
2. Model Training and Evaluation: Execute run.py for writing a submission file.
3. Hyperparameter Tuning: Use Cross_Validation_and_Tuning.ipynb to explore and fine-tune model parameters for optimal performance.