A Comparative Study on Demand Forecast of Car Sharing Users Based on ARIMA and LSTM

With the gradual growth of car-sharing market, it is particularly important to predict and analyze the demand of users. In this study, the K-means clustering method is used to cluster users into three types based on three indexes of the first use interval days, last use interval days, and monthly usage frequency. Each type of user demand has different time series trends. According to the ARIMA and LSTM models, respectively, the short-term demand of the three types of users is predicted. It is found that the LSTM model has a higher demand prediction accuracy for each type of user and the ARIMA model has a higher fitting accuracy. User demand forecasting can provide a reference for car-sharing managers to make scientific decisions.