EvanHolder/Tanzania-Well-Classifier

Flatiron Data Science - Module 3 Project # Module 3 Project - Predicting Functionality of Tanzanian Water Wells **Author**: Evan Holder **Flatiron Data Science**: Phase 3 Project **Github**: Link **Presentation**: Link # Repo Contents * README.MD * data * images * results_submissions * mod_3_well_functionality.ipynb - full analysis here # Introduction and Overview Water has been scarce in Tanzania for a long time. In recent years, a growing population, increased water consumption, and climate change have only exacerbated the water shortage. Accorinding to the World Health Organization, one in six Tanzanians lack access to safe drinking water. Many must travel miles on foot just to find water. Now, imagine you've spent the first two hours of the day walking in the sun, only to find a hour long lines, or worse, the tap is not functioning. In this project, I will be modeling Tanzanian well data from datadriven.org competition, *Pump it Up, Data Mining the Water Table*. The data includes information about the pump itself, and its operation. The goal is to predict which wells are functional, non-functional, and which need repair. Using the model, authorities will be able to better monitor and service pumps in need of repair and/or replacement. This model will help ensure that clean, drinkable water is available to more people across Tanzanian and help curb the water shortage. To build this model, I'll start with a decision tree. Next I'll make a random forest of trees, and finally I'll try out an XGBoost model. The model with the best accuracy will be the final model. ## Business Problem This project is all about increasing access to clean, usable well water across the country. So keeping that in mind, this project aims to increase the number of functional wells only. It does not aim to minimize cost, only to turn non-functional and functional needs repairs wells into functional wells. Additionally, the model shows which features are most important in determining well functionality. # Data Und …