基于相似日和PSO-Elman模型的共享单车需求量预测
Demand Prediction of Shared Bikes Based on Similar Days from the Perspective of Public Safety
周新民 1刘文洁 2闫志深 2胡怀钰3
作者信息
- 1. 湖南工商大学人工智能与先进计算学院,长沙 410205;湘江实验室,长沙 410205
- 2. 湖南工商大学计算机学院,长沙 410205
- 3. 湖南工商大学前沿交叉学院,长沙 410205
- 折叠
摘要
对各站点共享单车需求量的精准预测,能够提高共享单车管理与分配的效率,并有效防止供需失衡带来的公共秩序风险.通过综合考虑气象特征、时间特征以及历史数据对需求量的影响,提出了基于相似日和PSO-Elman神经网络模型.首先,研究时间特征对单车需求量的影响并对时间特征进行筛选,再通过皮尔逊相关系数验证并选择影响需求量的关键气象特征.随后采用灰色关联度算法,计算历史数据与待预测日的相似度并选取出相似日.最后,结合相似日数据和历史数据,将构建的PSO-Elman神经网络预测模型对高峰时段的单车需求量进行仿真预测.研究表明:与Elman及未综合考虑气象特征与时间特征的单车需求量预测模型相比,文章提出模型的预测结果具有更高的准确度.
Abstract
Accurate prediction of the demand for shared bikes at each station can improve the efficiency of shared bike management and distribution,and effectively prevent the risk of public order caused by the imbalance of supply and demand.By comprehensively considering the influence of meteorological characteristics,time characteristics and historical data on the demand,a model based on similar days and PSO-Elman neural network was proposed.Firstly,the influence of time characteristics on bicycle demand was studied and the time characteristics were screened.Then,Pearson correlation coefficient was used to verify and select the key meteorological characteristics affecting the demand.Then the grey relational degree algorithm was used to calculate the similarity between the historical data and the day to be predicted and select the similar day.Finally,combined with similar daily data and historical data,the PSO-Elman neural network prediction model was constructed to simulate and forecast the demand of bicycles in peak hours.The results showed that compared with Elman and the bicycle demand prediction model which did not consider the meteorological characteristics and time characteristics comprehensively,the prediction results of the proposed model had higher accuracy.
关键词
共享单车/需求量预测/相似日/灰色关联度/PSO-ElmanKey words
Sharing bikes/demand forecasting/similar day/grey correlation degree/PSO-Elman引用本文复制引用
基金项目
国家社会科学基金一般项目(21BGL231)
出版年
2024
NETL
NSTL
万方数据