Daniela-Toro/Pump-it-Up-Data-Mining-the-Water-Table

Intelligent water pump status prediction in Tanzania using machine learning. # 💧 Pump Guardian: Waterpoint Status Predictor **I built a machine learning model to predict the operational status of water pumps in rural Tanzania, using real-world data and Google Colaboratory.** ## 🧠 Project Overview This project was developed as part of the **"Pump it Up: Data Mining the Water Table"** competition on DrivenData. The goal was to classify each water pump into one of three categories: - **Functional** - **Non-functional** - **Functional needs repair** The dataset includes over 50 variables — such as geographic location, water quality, extraction type, and installer details — and comes with real-world challenges like: - Missing values - High-cardinality categorical features - Imbalanced class distribution ## 🚀 My Goals - Build and fine-tune a robust multi-class classification model - Handle missing data and complex categorical variables - Perform data cleaning, preprocessing, and feature engineering - Deliver explainable, reliable predictions ## 🧰 Tools & Technologies I worked entirely in **Google Colaboratory**, using the following tools and libraries: - `pandas`, `numpy`, `matplotlib`, `seaborn` - `scikit-learn`, `XGBoost`, `LightGBM` - `category_encoders`, `imbalanced-learn` - Grid search, cross-validation, and evaluation metrics ## 📁 Dataset All data was provided by the competition organizers and is available here: - 🔗 Competition overview - 🔗 Download the dataset - 🔗 Variable descriptions Files used: - `train.csv` – Features for training - `train_labels.csv` – Target variable - `test.csv` – Data for prediction - `SubmissionFormat.csv` – Required submission format > ⚠️ I followed the competition rules strictly and did **not** use any external data sources. ## 📈 Submission Details Submissions must include: - `id` (pump ID) - `status_group` (predicted class) Evaluated using the **macro F1-score**. - 🔗 Submit predictions ## 📌 Highlights - Hands-on application of machine learning to a real-world classification problem - Clear do …