Forecasting Bike Sharing Demand Using Machine Learning
This paper presents a model to forecast bike usage in San Francisco Bay Area bike sharing system using machine learning algorithms. Weather features are used to predict bike usage using two steps feature selection and model prediction. In feature selection process, linear Regression, ridge regression, lasso regression, recursive feature elimination, and random forest algorithms were used to select weather features. Gradient boost regressor and multi-linear perceptron regressor ﬁt on the selected important weather features. Multi-linear perceptron regressor outperforms gradient boost regressor by evaluating their F1 score 0.82 and 0.65.