jarretangbazo/pump-it-up

Classifying water pump functionality across Tanzania to support maintenance prioritization | DrivenData competition # Pump It Up ### Data Mining the Water Table **DrivenData Competition:** drivendata.org Predict which water pumps are faulty to promote access to clean, potable water across Tanzania. --- ## Table of Contents - Competition - Problem Definition - Task - Getting the Data - Data Files - Environment Setup - Project Structure - Local Setup - Tech Stack - Results and Key Findings - Results - Key Findings --- ## Competition ### Problem Definition Using data on water pumps in Tanzania collected by Taarifa and the Tanzanian Ministry of Water, the task is to classify each pump as **functional**, **functional needs repair**, or **non-functional**. Predictions draw on variables including pump type, installation date, geographic location, water source characteristics and management. The evaluation metric is **classification accuracy**. The training set has 59,400 observations and 41 features. **Why it matters** Nearly 57 million people in Tanzania rely on rural water infrastructure. A model that identifies failing pumps before communities lose access enables targeted maintenance, reduces downtime, and directs limited repair resources where they matter most. The decision this model supports: *which pumps should inspectors visit next?* ### Task Using data from Taarifa and the Tanzanian Ministry of Water, predict which pumps are functional, which need some repairs, and which don't work at all. | **Class** | **Train Count** | **% share** | |------|-------------|------------| | functional | 32,259 | 54.3% | | non-functional | 22,824 | 38.4% | | functional needs repair | 4,317 | 7.3% | ### Getting the Data 1. Sign up or log in at drivendata.org 2. Join the competition at the link above 3. Go to the **Data** tab and download all files into `data/raw/` #### Data Files | File | Description | |------|-------------| | `training_set_features.csv` | 59,400 pump records | | `training_set_labels. …