XGBoost-based 3-step feature learning is a combined embedded-wrapper method for feature selection, which integrates the feature importance ranking under various metrics, NSGA-II based metric selection and subset identification via recursive feature elimination. The method is targeting for supervised settings, and is testing on the NGSIM dataset for lane change intent prediction.
For more details, please find from paper: "A XGBoost-based lane change prediction on time series data using feature engineering for Autopilot vehicles", IEEE Transactions on Intelligent Transportation Systems (T-ITS), and please use the cite below:
Y. Zhang, X. Shi, S. Zhang and A. Abraham, "A XGBoost-Based Lane Change Prediction on Time Series Data Using Feature Engineering for Autopilot Vehicles," in IEEE Transactions on Intelligent Transportation Systems, doi: 10.1109/TITS.2022.3170628.Folder "1_NGSIM_Data_Cleaning_brief_Intro" gives a brief introduction on NGSIM dataset cleaning.
Folder "2_3Step_FeatureLearning_brief_Intro" provides the code for XGBoost-based 3-step feature learning algorithm.
Folder "experiment" provides the detailed codes for experiment in Section V D of paper "A XGBoost-based lane change prediction on time series data using feature engineering for Autopilot vehicles".