基于灰色—小波神经网络的有效泊位预测

发布时间：2018-01-15 11:44

本文关键词：基于灰色—小波神经网络的有效泊位预测　出处：《交通运输系统工程与信息》2017年05期 　论文类型：期刊论文

【摘要】：针对停车场有效停车泊位的变化特征,提出了基于灰色—小波神经网络的组合模型.先通过灰色单因素预测模型对有效停车泊位时间序列进行修正处理,再基于分步式小波神经网络模型对修正预测值进行运算,并通过马克科夫链预测模型得到更精确的预测区间,并利用实际案例分析,对模型的预测精度、稳定性、拟合度和训练时间进行了评价.研究表明,灰色—小波神经网络预测模型可降低初始数据波动性的干扰,与传统神经网络相比,预测结果误差波动性降低了10%~19%,稳定性提高了27%~33%,拟合度提高了10%~15%,精确度明显提高.
[Abstract]:A combined model based on gray wavelet neural network is proposed to deal with the time series of effective parking spaces by grey single factor prediction model. Then the modified prediction value is calculated based on the stepwise wavelet neural network model, and the prediction interval is obtained by the Markov chain prediction model, and the prediction accuracy of the model is analyzed by the actual case analysis. The stability, fitness and training time are evaluated. The results show that the grey wavelet neural network prediction model can reduce the disturbance of the initial data volatility, compared with the traditional neural network. The prediction results show that the error volatility is reduced by 10%, the stability is increased by 27%, the fitting degree is increased by 10%, and the accuracy is obviously improved.
【作者单位】：上海理工大学管理学院;
【基金】：国家自然科学基金(51608324) 上海市一流学科项目(S1201YLXK)~~
【分类号】：U491.7
【正文快照】： 0引言有效停车泊位是指当前停车场内可用来停放车辆的泊位[1],具有时间波动性、空间不均匀性、资源局限性等特征,采取合适的方法对有效停车泊位进行实时、精确预测,可使停车资源得到更合理的利用.对于有效停车泊位的预测研究,最先采用的是线性时间序列预测方法[2-3],该方法假

【相似文献】