In this study, a machine learning model was developed for predicting bike sharing demand using autogluon, a library for automated machine learning. A regression model was first trained on the original train dataset, then on a dataset with additional features, with hyperparameters optimization, and with model-specific hyperparameters optimization. The model trained on the dataset with additional features had a significant score improvement from 1.79702 to 0.57313 on Kaggle. Despite positive expectations about the hyperparameter optimization, the model had a worse score, but using a model-specific hyperparameter optimization improved the model score from 0.57313 to 0.53222. This result demonstrates that spending more time on hyperparameter optimization could improve the model even better.
For more information see: Report
Notebook: JupyterNotebook