摘要
旅游需求的精准预测对于旅游景区资源调度和管理有着重要作用。单一的浅层学习算法无法很好地拟合旅游客流量的特征,针对上述问题,本文通过组合深度学习和浅层学习算法,同时结合网络搜索行为数据,建立深度置信网络和利用自适应惯性权重优化后的自适应惯性权重优化的粒子群算法(APSO)去优化误差反向传播神经网络(BP)神经网络——APSO-BP的组合预测模型,用深度置信网络(DBN)模型对原始非线性客流量数据预测,再对DBN模型预测所产生的残差建立APSO-BP模型进行预测,将二者预测值合成后得到最终的预测值。通过实验证明了该组合模型能够较为准确的拟合客流量数据特征,有效地提高了预测精度。
The accurate prediction of travel demand is vital to the resources scheduling and management of the tourist attractions.A single shallow learning algorithm cannot well fit the characteristics of tourist traffic.According to the problem described above,this article combined deep learning and shallow learning algorithms,and internet searching behavior data to establish the combined prediction model of Deep Belief Network(DBN)and Back Propagation Neural Network(BP)optimized by Particle Swarm Optimization Algorithm with Adaptive Inertia Weight.This paper forecasted the originally nonlinear statistics of passenger flow volume through DBN model.Subsequently,the paper constructed PSOBP model to predict on the basis of residual generated from DBN model.Finally,the predicted values were added geometrically to obtain the final predicted values.It is predicted that the combined model can fit the data of passenger flow volume accurately and improve the prediction precise effectively.
作者
陆文星
戴一茹
李克卿
LU Wenxing;DAI Yiru;LI Keqing(School of management,Hefei University of Technology,Hefei 230009)
出处
《科技促进发展》
CSCD
2020年第5期470-478,共9页
Science & Technology for Development
基金
国家自然科学基金重点项目(71331002):基于云的管理信息系统再造研究,负责人:梁昌勇
国家自然科学基金重点项目(71771075):医疗结合的健康养老信息融合方法与云服务模式研究,负责人:梁昌勇
国家自然科学基金青年项目(71601061):智慧旅游环境下基于游客偏好的旅游云服务组合与推荐方法研究,负责人:赵树平
作者简介
陆文星,副教授,硕士生导师,研究方向:项目管理、信息管理和信息系统、决策分析等;通讯作者:戴一茹,在读硕士研究生,研究方向:信息管理与信息系统、算法优化、数据分析等。