Prediction of Shared Bicycle Entry and Exit Flow based on LSTM Model

As the use of shared bicycles becomes increasingly prevalent in urban areas, the issue of short-distance travel for residents has been significantly alleviated. Moreover, within the 30-minute radius of daily life, bicycles serve as a convenient solution for short-distance travel. However, the deployment strategy employed by shared bike operators lacks scientific rationality. The placement of bicycles in inappropriate locations compromises their convenience. This research aims to establish a predictive model for forecasting the entry and exit flow of shared bicycles in each region of the city. The shared bicycle data for January 2024 were collected from the Divvy online platform in Chicago. Python was utilized to extract and process the trip data, and time features were handled by minute-wise aggregation of entry and exit flows. By selecting data from the past 30 minutes as input features, normalizing the data, and constructing time series data suitable for LSTM input, the predictive model was developed. Following model training, traffic prediction was conducted using the test set, and the model's performance was evaluated. Ultimately, the effectiveness of the model was intuitively understood by plotting comparative graphs between actual and predicted values. This research aims to provide real-time decision support for urban traffic management and shared bike operations by offering insights into the predictive trends.