brendanvaz/Exploratory-Evaluation-of-Water-Pumps-In-Tanzania

Exploratory Evaluation of Water Pumps in Tanzania Overview This group project analyzes the operational status of water pumps in Tanzania using machine learning techniques. The study was led by me (Brendan Vaz), with team members collaboratively working on different aspects of the project. Leveraging the "Pump it Up: Data Mining the Water Table" dataset from the Taarifa waterpoints dashboard, this analysis provides actionable insights into pump functionality, maintenance needs, and factors influencing failures. The project evaluates multiple machine learning models to identify the best performer for predicting water pump functionality. Objectives To predict the functionality of water pumps (functional, functional but needs repair, non-functional) using machine learning models. To analyze key factors influencing water pump performance. To recommend data-driven strategies for maintenance and resource allocation. Dataset Source: Pump it Up: Data Mining the Water Table Description: This dataset includes 38 features related to water pump functionality, such as: Pump type Management authority Geolocation Installation date Water quality and quantity Methodology Leadership Role: As the project leader, I (Brendan Vaz) coordinated all aspects of the project, from defining objectives to final delivery. I facilitated team collaboration, assigned responsibilities, and ensured deadlines were met. Personally oversaw critical components such as data preprocessing, model evaluation, and documentation. Data Preprocessing Imputed missing values using MICE (Multivariate Imputation by Chained Equations) for numerical features. Removed redundant or highly correlated columns to improve model performance. Encoded categorical variables using one-hot encoding. Feature Engineering Created derived features like pump age to improve predictive performance. Model Training Tested the following machine learning models: Decision Trees Random Forest Support Vector Machines (SVM) K-Nearest …