Student Dropout Prediction Pipeline

This repository contains a machine learning pipeline for predicting student dropouts. The main notebook, dropout_pipeline.ipynb, guides you through the entire process — from data preprocessing to model evaluation and prediction.

Repository Structure

• dropout_pipeline.ipynb: The main Jupyter notebook that contains the complete pipeline for predicting student dropouts.
• data/: Directory containing the dataset used for training and evaluation.
  • history_engcomp_and_ccomp_all_semesters_until_2023_1.csv: The dataset containing historical student data.
• models/: Directory containing the trained machine learning models.
• predictions/: Directory containing the prediction results.

Notebook Overview

1. Importing Libraries

The notebook begins by importing the necessary libraries for data manipulation, visualization, and machine learning.

2. Data Loading and Preprocessing

• Load the dataset from data/history_engcomp_and_ccomp_all_semesters_until_2023_1.csv.
• Perform data cleaning and preprocessing to prepare the dataset for modeling.

3. Exploratory Data Analysis (EDA)

• Visualize the distribution of features and the target variable.
• Generate insights to inform the modeling process.

4. Model Training, Tuning, and Predictions

• Train multiple machine learning models, including Random Forest, XGBoost, and LightGBM.
• Save the trained models in the models/ directory.
• Evaluate model performance using the accuracy metric.
• Compare the performance of different models to identify the best-performing one.
• Use the trained models to make predictions on new data.
• Save the predictions to the predictions/ directory.

Results

The prediction results are saved in the predictions/ directory. The notebook includes visualizations and evaluation metrics to document the performance of the models.