Water Pump Classification
Description
A DrivenData.org machine learning competition to classify water pumps in Tanzania as functional, functional in need of repair, or non-functional
Goal
Acheive highest possible classification rate on unseen test data.
Outcome
Best submission acheived a classification rate of .8252
Notable Features
Best submission ranks in the top 1.5% of submissions to date.
Technology Used
- Python
- scikit-learn
- XGBoost
- Random Forest
- Bagging Classifier
- K-Nearest Neighbors
- Target encoding
- Iterative imputation
- Grid search, Bayesian hyperparameter optimization