Research on the Forecast of Shared Bicycle Rental Demand Based on Spark Machine Learning Framework

In recent years, the shared bicycle project has developed rapidly. In use of shared bicycles, a great deal of user riding information is recorded. How to extract effective knowledge from these vast amounts of information, how to use this knowledge to improve the shared bicycle system, and how to improve the user experience, are problems to solve. Citi Bike is selected as the research target. Data on Citi Bike’s user historical behavior, weather information, and holiday information are collected from three different sources, and converted into appropriate formats for model training. Spark MLlib is used to construct three different predictive models, advantages and disadvantages of different forecasting models are compared. Some techniques are used to enhance the accuracy of random forests model. The experimental results show that the root mean square error RMSE of the final model is reduced from 305.458 to 243.346.