This very complicated machine learning project is based on a seismic dataset https://www.kaggle.com/c/LANL-Earthquake-Prediction. Aiming to predict the time remaining before the next laboratory earthquake, the project is a combined application of complex feature engineering of time series, ensemble learning of "xgboost", "lightgbm" and "catboost", and a parameter turning package "hyperopt". Besides those techniques, NN, CNN, RNN, LSTM and the "gplearn" package are under test. Note that the main function is in seismic_prediction.py, and a lot of settings are also there.
https://www.kaggle.com/artgor/earthquakes-fe-more-features-and-samples
https://www.kaggle.com/scirpus/andrews-script-plus-a-genetic-program-model