Online booked vehicle flow prediction method based on deep learning

The invention discloses an online car-hailing flow prediction method based on deep learning, and belongs to the field of machine learning research. According to the method, a two-way long-short-term memory network model with historical traffic volume as input is established, that is, data of past 8-24 hours are used as input, and car-hailing flow data of the next hour are predicted. The method includes the following steps: (1) acquiring online car-hailing operation historical data, and counting online car-hailing traffic flow; (2) carrying out flow change analysis on the online car-hailing operation data, and matching and marking external environment attributes of sudden increase or sudden decrease of the flow; (3) carrying out DBSCAN clustering; (4) expanding a clustering region according to the information points; and (5) constructing an LSTM (Long Short Term Memory) prediction model taking historical traffic volume distribution as input, and predicting online booked traffic flow data of each region in a fu