Congratulations for all winners. It was fantastic challenge.

The Solution Steps:

Architecture Diagram & ETL Process :

Extract : Data is sourced from CSV files hosted on GitHub, including training data, test data, sample submissions, and variable definitions.

Transform : Involves preprocessing, handling missing values, feature engineering (including geospatial transformations using H3 hexagons, distance calculations, and encoding categorical variables

Load : Transformed data is temporarily stored in Pandas DataFrames and used directly for training machine learning models.

Data Modeling :

• Utilizes feature engineering techniques like geospatial processing with H3 hexagons and distance calculations.
• Multiple algorithms are applied, including CatBoost and HistGradientBoost.
• Model performance is evaluated using Root Mean Squared Error (RMSE).

Inference :

The trained model predicts outcomes on test data, outputting two columns, including the target variable.

Performance Metrics

Evaluation is based on RMSE, with KFold cross-validation ensuring reliability.

Public score: 3.47; Private score: 3.50.

for more details about the full solution please visit:

Yousifshaheen/International-Women-s-Day-Challenge-In-Zindi-Solution: The goal of the challenge is predict women headed households living below an income threshold