Research on Deep Learning Model for Location of High-speed Railway Freight Stations
The rapid growth of express delivery volume in recent years has resulted in the high-speed railway(HSR)ex-press mode.However,in the planning process,there is a complex heterogeneous and non-linear relationship between the site selection of HSR freight stations and logistics demand and facility conditions.In this context,this study developed a deep learning-based method for selecting the location of HSR freight stations.By initially constructing the site selection model based on three dimensions:market demand,geological location and transportation,and development environment,as well as 10 basic features and deep learning algorithm,this study optimized the hyperparameters by using genetic algorithm,and tested the model according to the importance of each feature in order to determine the best feature system and model learning layer.The experimental results show that the feature subset composed of station node cover-age,high-speed train reservation,urban express delivery volume,road traffic accessibility,and the number of surround-ing airports shows the best site selection performance in the deep learning model stacked with LSTM and Dense,with F1 score up to 0.944 4,a significant improvement compared to that before optimization.This paper further takes Zhejiang province as an example to empirically analyze the optimized deep learning model,and the site selection results also con-firm that the deep learning model established by the study learns the actual HSR freight station site selection law suffi-ciently and can help to make more scientific and structured HSR freight station site selection decisions.
万方数据
维普
