shammy-lp/tanzania-water-churn

# **Tanzania Water Well Condition Prediction** ## **Project Overview** Tanzania, a developing country with a population of over 57 million, faces challenges in providing clean water to its citizens. While thousands of water wells exist, many are **non-functional** or require **urgent repairs**. This project aims to **predict the condition of a water well** based on key factors such as **pump type, installation year, and geographic location**. By building a **machine learning classifier**, we can help NGOs and government agencies **prioritize well repairs, allocate resources efficiently, and plan for better water infrastructure**. --- ## **Problem Statement** The dataset contains records of water points across Tanzania, including their operational status. The goal is to classify each well into one of three categories: - **Functional** – The well is fully operational. - **Functional but needs repair** – The well is working but requires maintenance. - **Non-functional** – The well is broken and needs full replacement. --- Pump It Up: Data Mining the Water Table - Data Source ## **Dataset Description** The dataset includes various features that influence a well's condition, such as: - **Geographical Data:** Region, district, latitude, and longitude. - **Well Characteristics:** Pump type, extraction type, water source, and management type. - **Operational History:** Installation year, recorded usage, and payment method. --- ## **Data Preparation & Preprocessing** To ensure a high-quality model, the following preprocessing steps were performed: **Handling Missing Values** – Imputing missing data using median (for numerical features) and most frequent (for categorical features). **Feature Engineering** – Extracting useful information, such as the installation year difference. **Encoding Categorical Variables** – Using **One-Hot Encoding** for categorical variables. **Scaling Numerical Features** – Using **StandardScaler** for normalization. **Train-Test Split** – Sepa …