Multivariate Passenger Flow Forecast Based on ACLB Model

Abstract With the rapid increase in urban population, traffic problems are becoming severe. Passenger flow forecasting is critical to improving ability of buses meet travel needs residents and alleviating pressure. However, factors affecting passenger have complex non-linear characteristics, which creates a bottleneck prediction. Deep learning models CNN, LSTM, BISTM gradually emerging attention mechanism key points solve above problems. Based on summarizing characteristics various models, this paper proposes multivariate prediction model ACLB extract nonlinear spatio-temporal data. We compare performance with BILSTM, CNN-LSTM, FCN-ALSTM through experiments. better than other models.