The code is for predicting the severity of the impact of an earthquake on a building. Classification is performed through several experiments, including using models like Logistic Regression, SVM, XGBoost, Neural Networks, and Random Classifier; with the highest configuration search using Grid Search and Randomized Search; as well as based on the highest feature correlations (with an experimentally adjustable THRESHOLD) as the main predictor features of the model.
This code successfully achieved fourth place on the public leaderboard of the ML Olympiad 2024.
You can cite the olympiad/competition in:
@misc{ml-olympiad-predicting-earthquake-damage,
author = {Tensor Girl},
title = {ML Olympiad - Predicting Earthquake Damage},
publisher = {Kaggle},
year = {2024},
url = {https://kaggle.com/competitions/ml-olympiad-predicting-earthquake-damage}
}