JimmyHAL/Tanzania-Waterwell-Classification

A classification project determining the condition of water sources in Tanzania using the data release by the goverment. # Tanzania Water Point Classification The purpose of this project is to take data provided by the Tanzania goverment about numerous water source points in the country and create a classifier that categorize new waterpoint data into three categories which are - **'functional'** - waterpoint is operational **'non functional'** - waterpoint is not operational and needs repair **'functional needs repair'** - waterpoint is operational but needs maintainence repair ### Repository Files The following are folders and files you can find in this repository and their discription 1. content 1.training_set_labels - The dataset target 2.training_set_values - The dataset features 3.test_set_values - The features for prediction 4.column_descriptions.txt - Discription of the features 2. images - images used in this README.md 3. README.md - discription of the project and repository 4. presentation.pdf - pdf of a powerpoint presentation meant for non-technical audience ### Process For this project I followed OSEMN framework of- 1. Obtain the data 2. Scrub the data 3. Explore the data 4. Model the data 5. iNterpret the data ## Obtain - The Dataset The dataset is from drivendata.org. It is provided by Taarifa and the Tanzanian Ministry of Water. The raw data contains 40 features columns and 3 categories for the target. ## Scrub For this project I made heavy use of the pandas library to manipulate and explore the dataset. I also made heavy use of sklearn libraries to model, evaluate and test the data. The first thing I did is use pandas profiling to explore the data. I found that there were several highly correlated columns and some columns missing a massive amount of data. I also made use of mosaicplot to look at several columns including the 'payment' column We can see that pay annually has the most functional waterpoints, while never pay and unknown has the least. After taking a look at the features description …