摘要
针对如何利用消费者购买行为数据进行产品销量的有效预测问题,以高卷入度产品——汽车为例,挖掘用户生成的评论以及搜索行为留下的数据,构建基于不同数据来源的预测模型,采用线性以及非线性机器学习算法对不同的预测模型求解均方误差MSE。研究结果表明,增加用户关注度的数据能够增强基准数据的预测效果,而添加具体口碑数据后,可以有效地增强搜索数据的预测效果。
To solve the problem of how to use the data of consumers'purchasing behavior to predict the sales volume of products effectively,the high-involved product automobile is taken as an example,the data left by the user generated comments and search behavior were mined,and prediction models were built based on different data sources,linear and nonlinear machine learning algorithm were used to solve the MSE of different prediction models.The results show that the prediction effect of benchmark data was enhanced by the search data,and the prediction effect of search data can be effectively enhanced by the data of adding specific word-of-mouth data.
作者
章金成
张宁
张锐
ZHANG Jin-cheng;ZHANG Ning;ZHANG Rui(School of Business, Qingdao University,Qingdao 266100, China)
出处
《青岛大学学报（自然科学版）》
CAS
2020年第2期58-62,共5页
Journal of Qingdao University(Natural Science Edition)
基金
国家民委民族研究项目(批准号:2018-GMB-022)资助
山东省社科规划项目(批准号:18CHLJ22)资助
青岛市社科基金(批准号:QDSKL1601077)资助。
关键词
网络口碑
搜索指数
销量预测
机器学习
word-of-mouth
search trend data
sales forecasting
machine learning
