antonisleventakis / predicting-student-final-grades-with-regression-models

Regression Machine Learning Pipelines for predicting student performance

Predicting-Student-Final-Grades-with-Regression-Models

Description: This repository contains two different implementations of regression models for predicting student final grades. The first model is a neural network, while the second model is a semi-supervised algorithm.

The neural network model is built using TensorFlow, a popular open-source machine learning library. It takes as input various features related to student performance, such as previous grades, attendance, and study habits. The model is trained using a large dataset of student records, and then tested on a held-out set of data to evaluate its accuracy. The code includes detailed comments to help users understand the structure and operation of the model.

The second model is a semi-supervised algorithm that uses both labeled and unlabeled data to make predictions. It leverages the fact that some student data may be missing or incomplete, and uses clustering techniques to group similar students together. It then assigns labels to these groups based on the known final grades of some students, and uses this information to make predictions for the remaining students. The code includes several helper functions and visualization tools to help users explore the data and understand the algorithm's output.

Both models are designed to be flexible and easy to use, with options for adjusting hyperparameters and input data. They can be run locally or in a cloud environment using popular platforms such as Google Colab. The code is fully documented and available under an open-source license, allowing other researchers and developers to build on this work and contribute to the field of student performance prediction.