The bike-sharing service system is a service that allows a customer to rent a bike from a bike-sharing station and then return it to another bike-sharing station in a short time after they reach their destination. Thus, the impact of the bike distribution system based on the frequency of bike usage needs to be assessed. The bike-sharing system operator needs to predict the demand to accurately know how many bikes are needed in every station so as to assist the planner in the management process of the bike-sharing stations. This paper proposes an efficient and accurate model for predicting the bike-sharing service usage using various features of a machine learning algorithm. We compared the exiting techniques for the sequential data predicting of artificial intelligence for time series data and analysis. Then, we considered the use of the multivariate model with a recurrent neural network (RNN), a long short-term memory (LSTM), and a gated recurrent unit (GRU). In addition, we considered combining the LSTM and GRU methods together to improve the model’s effectiveness and accuracy. The results showed that all the RNNs, including the LSTM, GRU, and the model combining the LSTM and GRU, are able to achieve high performance using the mean square mean absolute, mean squared error, and root mean square error. However, the mixed LSTM–GRU model accurately predicted the demand in this case.