tarunk42/Multiclass-Classification-for-Predicting-Faulty-Water-Pumps-in-Tanzania

# Multiclass Classification for Predicting Faulty Water Pumps in Tanzania This project aims to develop a predictive model for identifying the status of water pumps as functional, in need of repair, or non-functional based on various features related to the pumps’ characteristics and management. The study utilizes a comprehensive dataset containing 40 training features and 1 training label, with 30 categorical and 10 numerical columns. Exploratory data analysis and preprocessing techniques are applied to gain insights into the data’s completeness, handle missing values, and address feature imbalances and redundancies. Additionally, the CatBoost and Logistic regression algorithm compared. The research also investigates the relationship between the functionality status of water pumps and the population of the areas in which they are installed. Furthermore, it identifies key factors influencing the functional state of the pumps, offering valuable insights for improving the availability of potable water to communities. The results of the study show that the logistic regression model achieves a commendable accuracy of 74% on the test set. However, CatBoost, with hyperparameter tuning, outperforms logistic regression, achieving an accuracy of 81.15%