FridahKimathi/Water-Pump-Functionality-Prediction-in-Tanzania

The project uses a supervised machine learning model to predict with high accuracy whether a pump is functional or not. This has the potential to greatly improve access to clean water in Tanzania and other developing countries. # Water-Pump-Functionality-Prediction-in-Tanzania ##### **Author**: Fridah Kimathi ## Overview **** The project aims to develop a model to classify the functionality status of water pumps in Tanzania using data sourced by Taarifa and the Tanzanian Ministry of Water . ## Business Problem *** Tanzania is a developing country which is in the midst of a water crisis. The country struggles to provide clean water to its population of over 57 million people despite the fact that the country has many established water points. The Tanzanian government through the Ministry of Water aims at resolving the water crisis in the country by maintaining and repairing the water pumps in time. In order for its Engineers to achieve the objectives faster, they need to be know in advance which water pumps are likely to fail and understand the causes of failure. The project therefore, aims at developing a model using data collected by Taarifa and the Tanzanian Ministry of Water to predict the functionality status of the water pumps in Tanzania as well as provide important insights on the main factors contributing to water pump failure in Tanzania. The model created will enable the Tanzanian Ministry of Water to improve the maintenance operations its water pumps. ## Data *** The data used in this project is from the Pump it Up: Data Mining the Water Table competition hosted by DrivenData, originally sourced by Taarifa and the Tanzanian Ministry of Water . ## Modelling *** Different models were evaluated and the best performing model was picked to be the final model.The Random Forest Classifier was picked as the final model, with its parameters being the best parameters found through grid search. Models ## Evaluation *** The accuracy score of the model is 0.7937. That means that the model predicts the correct water pump functionality status 79.37% of the time. The most important features according to the final model are amount_tsh, permit, public_meeting, distri …