sgchatrathi/tanzania-water-projectp3

# Tanzania Well Functionality: A Management Perspective By: Samira Chatrathi and Brian Reynolds # Overview In this project we take Tanzanian Well Data from the DrivenData competition to build a classification model to predict which wells are functional, non functional, and functional and need repair. # Business Understanding/Stakeholder Tanzania has struggled with finding a solution to their water crisis for years. Not only is the affordability of clean water an issue, but the infrastructure and maintenance for water has also proven to be a consistent problem for Tanzania. As a means of supporting Tanzania's sustainable development, this ongoing issue needs to be fundamentally reevaluated. Our stakeholder is Water.org, a non-government organization that is looking to implement a vast manageable amount of water pumps within the country. Before deploying teams into Tanzania blindly, they want to see how they can optimize their time and resources to effectively predict whether a water pump is functional, non existent, or just broken and needs repair. Not only is Water.org focusing on fixing non functional wells themselves, but they are also focusing on prioritizing efficiency and optimizing long term scalability. # Data The dataset includes information on 59,400 wells, each of which has 41 different features. We also implemented spacial data to see how certain physical characteristics could better support our model overall. In our EDA, we were able to see that 46% of the total wells in Tanzania were either non-functional, or broken and in need of repair. We broke up our data into two columns to further understand our analysis. We handled missing values and used SMOTE to address class imbalance within our modeling stage. Our features were separated into Physical and Politcal. This technique helped us understand where actionable insights can be associated with our features. # Modeling After using our dummy classifier model, we were able to implement a decision t …