论文部分内容阅读
通过分析在线拍卖出价特点,利用决策树和Bagging算法建立了一种全新的在线拍卖成交价格预测模型.作者编写程序收集淘宝网在线拍卖交易数据3310条,对应有效出价记录8275条.数据分析表明,如不考虑未成交商品,则有40.4%的交易可以利用出价次数精确计算最终成交价格.如将未成交商品视为成交价格为0,该比例可提高为79.55%.据此发现,作者通过预测出价次数间接对成交价格进行预测.实验证明,模型明显优于平均值预测,并有21.7%的预测结果完全准确.通过与Heijst发表于《Decision Support Systems》上的研究进行对比,结果表明预测模型在样本需求量、运算时间,及完全准确预测率上有明显优势.由于模型训练时间仅为数秒,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.
By analyzing the characteristics of online auction bids and using decision tree and Bagging algorithm to establish a new online auction transaction price forecasting model. Author program to collect online auction data Taobao 3310, corresponding to the effective bidding record 8275. Data analysis shows that, If you do not consider the unfinished goods, 40.4% of the transactions can use the number of bids to accurately calculate the final transaction price.If the unfinished goods as the transaction price is 0, the ratio can be increased to 79.55% .From this discovery, the author through the forecast The number of times of bidding is used to predict the transaction price indirectly.Experiments show that the model is superior to the average forecast and 21.7% of the forecast results are completely accurate.Comparing with Heijst’s research on “Decision Support Systems”, the results show that the forecasting model There is a clear advantage in sample demand, computing time and completely accurate forecasting rate.As the model training time is only a few seconds, it lays a foundation for establishing real-time online auction price forecasting decision support system.