基于深度学习的有效停车泊位预测模型  

Prediction Model of Effective Parking Space Based on Deep Learning

作　　者：温浩宇[1] 赵灵君 王帆 于江霞[1] WEN Haoyu;ZHAO Lingjun;WANG Fan;YU Jiangxia(School of Economics and Management,Xidian University,Xi'an 710126,Shaanxi,China;Xi'an Big Data Asset Management Co.,Ltd.,Xi'an 710075,Shaanxi,China)

机构地区：[1]西安电子科技大学经济与管理学院,陕西西安710126 [2]西安大数据资产经营有限责任公司,陕西西安710075

出　　处：《重庆交通大学学报（自然科学版）》2022年第2期30-34,43,共6页Journal of Chongqing Jiaotong University(Natural Science)

基　　金：陕西省自然科学基金项目(2020JM-211);教育部人文社会科学规划基金项目(20YJAZH123);西安市科技局软科学项目(XA2020-RKXYJ-0143)。

摘　　要：针对传统有效停车泊位预测方法无法刻画泊位前后时刻关联关系的问题,采用基于深度学习的LSTM(long short-term memcry)神经网络对其进行改进,提出了LSTM有效停车泊位预测模型,并基于此模型对不同类型的停车区域进行分析与预测。在构建模型的基础上,综合考虑了有效停车泊位预测的时空特性,选取目标区域内多个邻近停车场的历史停车数据组成数据集,并构建有效停车泊位预测的对比模型,以此检验模型的预测精度。研究结果表明:在不同类型停车区域的有效停车泊位预测中,LSTM模型预测结果与真实值一致性较好,预测精度均高于BP预测模型和ARIMA预测模型;LSTM模型在有效停车泊位预测方面可靠且有效。Aiming at the problem that the traditional effective parking space prediction method could not describe the time correlation before and after the parking, LSTM neural network based on deep learning was given to improve traditional prediction methods, and a prediction model of effective parking space based on LSTM was proposed to analyze and predict different types of parking areas. Based on the established model, the temporal and spatial characteristics of effective parking space prediction were comprehensively considered, the historical parking data of multiple adjacent parking lots in the target area were selected to form a data set. The comparison model of effective parking space prediction was constructed to test the prediction accuracy of the model. The results show that in the prediction of effective parking spaces in different types of parking areas, the prediction results of the proposed model are in good agreement with the real value, and its prediction accuracy is higher than that of BP prediction model and ARIMA prediction model. It is indicated that the proposed model is reliable and effective in the effective parking space prediction.

关 键 词：交通运输工程 静态交通 停车泊位预测 深度学习 LSTM神经网络 

分 类 号：U491[交通运输工程—交通运输规划与管理]