Using data mining techniques for bike sharing demand prediction in metropolitan city

Abstract

Currently Rental bikes are introduced in many urban cities for the enhancement of mobility comfort. It is important to make the rental bike available and accessible to the public at the right time as it lessens the waiting time. Eventually, providing the city with a stable supply of rental bikes becomes a major concern. The crucial part is the prediction of bike count required at each hour for the stable supply of rental bikes. A Data mining technique is employed for overcoming the hurdles for the prediction of hourly rental bike demand. This paper discusses the models for hourly rental bike demand prediction. Data used include weather information (Temperature, Humidity, Windspeed, Visibility, Dewpoint, Solar radiation, Snowfall, Rainfall), the number of bikes rented per hour and date information. The paper also explores an filtering of features approach to eliminate the parameters which are not predictive and ranks the features based on its prediction performance. Five Statistical regression models were trained with their best hyperparameters using repeated cross-validation and the performance is evaluated using a testing set: (a) Linear Regression (b) Gradient Boosting Machine (c) Support Vector Machine (Radial Basis Function Kernel) (d) Boosted Trees, and (e) Extreme Gradient Boosting Trees. When all the predictors are employed, the best model Gradient Boosting Machine can give the best and highest R2 value of 0.96 in the training set and 0.92 in the test set. Furthermore, several analyzes are carried out in Gradient Boosting Machine with different combinations of predictors to identify the most significant predictors and the relationships between them.