A subway passenger flow prediction method based on stl-lstm

Abstract

The invention relates to a subway passenger flow prediction method based on STL-LSTM. First, the subway passenger traffic data in a statistical period is calculated; the STL addition model is used to decompose the subway passenger traffic data to obtain three decomposition sequences; then, the LSTM is set parameters to build the LSTM prediction model. Train the LSTM model with the same parameters on the 3 decomposed sequences and make the test set predictions. Then, the prediction results of the test set of the three decomposition sequences are added to obtain the prediction results of the subway data test set, and the average relative error is calculated; the decomposition period and time step are modified, and the above process is repeated, and the modification is stopped when the average relative error is basically stable; , use the model with the smallest average relative error in the test set to predict the passenger volume after the prediction steps of the three sequences obtained, and add the results to obtain the final prediction result. The present invention can improve the accuracy of subway passenger flow prediction.

Description

technical field [0001] The invention relates to the field of subway passenger flow prediction, in particular to a subway passenger flow prediction method based on STL-LSTM. Background technique [0002] Subway passenger flow is an important basis for subway operation and management. Accurate prediction of subway passenger flow in the future is of great significance for subway operation departments to adjust travel plans. Considering the obvious periodicity, trend and randomness of subway passenger flow, it should be reflected in the establishment of prediction model. [0003] Decomposing the time series can eliminate the influence of periodic factors and reflect the real objective laws and trends of the time series. In recent years, many researchers have carried out in-depth research on it. Time series decomposition models can be divided into X11 series, X12 series, SABL, wavelet analysis and EMD according to different strategies. These models can decompose the time serie...

AI Technical Summary

Problems solved by technology

STL decomposes passenger flow into periodic sequence, trend sequence and random sequence, but the trend of subway passenger flow data is not obvious, and the periodicity of subway passenger flow data is mainly weekly. The effectiveness of STL for subway passenger flow data has yet to be verified.

Images