Open AccessTandem Prediction Research on LSTM-DT
Author(s) -
Min Wang,
Xin Liu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1769/1/012038
Subject(s) - computer science , generalization , tandem , interval (graph theory) , decision tree , tree (set theory) , data mining , time series , artificial intelligence , machine learning , mathematics , mathematical analysis , materials science , combinatorics , composite material
In order to effectively avoid the problem of bus tandem, the factors and prediction model of bus tandem based on multi-source data are studied. Due to the problems of complex attributes and low prediction accuracy, this paper presents a prediction method based on long short time memory and decision tree (LSTM-DT). Based on the existing historical GPS data, a variety of input data schemes are formed by different combinations of static spatial data and dynamic spatial data. In the process of tandem prediction, the long short time memory (LSTM) model is used to predict the time series characteristics of stations with different intervals, and the attribute and weight are adjusted by combining with decision tree. Through the LSTM-DT model, it gives the comparison of the tandem prediction in different time periods and different station intervals. The experimental results show that the LSTM-DT prediction model has better generalization ability than the traditional model in the case of more input factors, and has higher prediction accuracy for the tandem at a certain interval.