基于支持向量机的在线拍卖成交价格预测方法  

A support vector machine based method for predicting final prices of online auctions

作　　者：刘洋[1] 冯玉强[1] 邵真[1] 

机构地区：[1]哈尔滨工业大学信息系统研究所,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2010年第3期400-404,共5页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目(70572023)

摘　　要：针对现有在线拍卖成交价格预测方法计算时间长的问题,依据淘宝网在线拍卖出价特点,建立了一种全新的在线拍卖成交价格预测方法.方法以支持向量机分类算法为基础,通过预测出价次数间接对成交价格进行预测.利用编写程序收集淘宝网在线拍卖交易数据3 310条,对应有效出价记录8 275条,以之作为实验数据.实验证明,预测结果明显优于平均值预测,并有22.1%的预测结果完全准确.由于训练时间仅为数秒,为建立实时在线拍卖成交价格预测决策支持系统奠定了基础.A new method was proposed for predicting the final price in online auctions. By analysing bidders behavior, the problem of long calculation time was solved. Instead of predicting the final price directly, the method used a support vector machine to analyse the time of each bid and then used that data to calculate an end price. The authors collected data on 3310 transactions and their corresponding 8275 bids from Taobao and used them as experimental data. The experiment proved that this model substantially outperforms the naive method of predicting by using the mean category price. The training time was only a few seconds, yet 22.1% of the predicted results were identical to the real ones. This research can provide a foundation for the development of real-time decision support systems.

关 键 词：在线拍卖 成交价格预测 支持向量机 

分 类 号：C931.9[经济管理—管理学]