tj-murphy/water-pump-functionality

Machine learning system developed to predict the operational status of water pumps in Tanzania, addressing critical infrastructure challenges in developing regions. Predicting Water Pump Functionality in Tanzania - This repository contains the code and documentation for the project "Predicting Water Pump Functionality in Tanzania using Machine Learning Techniques". The project aims to address the critical issue of non-functional water pumps in Tanzania by developing machine learning models to predict the operational status of these pumps. The project was conducted from January 2024 to May 2024. Project Overview: - Access to clean water is a significant issue in Tanzania, with many water pumps being non-functional or in need of repair. This project utilises a comprehensive dataset that contains various features related to water pumps, including geographical information, pump characterisitcs and management details. The primary goal is to develop predictive models using multiple machine learning techniques and compare their performance. Repository Contents: - - "report.pdf" : Detailed project report covering the literature review, methodology, results, and discussion. - "CW2-final.ipynb" : Jupyter Notebook containing the code for data analysis, preprocessing, and model development. Models Used: - 1. K-Nearest Neighbours (KNN) 2. Random Forest 3. XGBoost 4. CatBoost 5. TabNet Results: - The performance of each model was evaluated using precision, recall, F1 score, and accuracy. The gradient boosting algorithms (XGBoost and CatBoost) showed the best performance, achieving an accuracy of 79%. The detailed results and comparisons are available in the project report ("report.pdf"). Data: - The dataset used in this project is provided by the Tanzanian Ministry of Water in collaboration with Taarifa. It includes information on 59,400 water pumps with 41 features. The dataset is not included in this repository and should be downloaded from the DrivenData competition page: drivendata.org Grade: - This project received a grade of 78%. Acknowledgements: - This project was completed as part of the C …