Diabetes Prediction Machine Learning Project

Overview

This repository contains a machine-learning project focused on predicting diabetes using relevant medical data. This project aims to develop a model that can accurately predict the likelihood of an individual having diabetes based on various input features.

Dataset

The dataset used for this project consists of 2460 samples with 8 features, including age, BMI, insulin, and more. The dataset was obtained from Kaggle.

Approach

In this project, I employed several different machine learning models from Scikit-learn to train and evaluate the predictive model. The steps involved in the process include:

• Importing data: Cleaning the dataset, handling missing values, and scaling features.
• Understanding data: Understanding data distribution and correlation matrix
• Model training: Training the selected algorithm on the preprocessed dataset.
• Model evaluation: Assessing the performance of the trained model using appropriate metrics such as accuracy, precision, recall, F1-score, and ROC-AUC.

Results

Currently, the trained model achieved an accuracy score of 80%.

Usage

To run this project locally, follow these steps:

1. Open this link to access the Google Colab file and make a copy
2. Download dataset files from this repository
3. Run the Google Colab to preprocess the data, train the model, make predictions, and save the model.

Technologies Used

Programming languages: Python

Libraries: Scikit-learn Pandas NumPy Matplotlib Seaborn

Future Improvements

Some potential enhancements for this project include:

• Exploring additional features or data sources to improve the model's predictive performance.
• Tuning hyperparameters to optimize the model's performance.
• Deploying the trained model as a web application or API for real-time predictions.

Acknowledgments

Special thanks to Ehab Aboelnaga for providing the dataset used in this project.