基于时间序列模型商品搜索排序

SORTING OF COMMODITY SEARCH BASED ON TIME SERIES MODEL

电商商品搜索引擎目前大多都是基于用户浏览行为与购买行为来建立搜索因子,通过这些因子计算出商品分地区的排序分值进行排序。然而这些行为数据都是属于历史数据。搜索的结果都是基于历史数据的分值计算没有前瞻性。特别对于换季商品使用该方式计算出来的搜索结果不佳,转化率不高。提出一种基于时间序列的分析方法,对部分的搜索因子采用预测数据来计算分值,以满足商品搜索中某些具有周期性季节性商品的合适排序。

Commercial search engine which establishes search factors are mostly based on the current user browsing behavior and buying behavior, and through these factors to calculate the goods sorted by region sorting scores. However, these behavioral data belong to historical data. The search results based on the historical data of the score calculation is not forward-looking. Especially for the seasonal goods, the method of calculation of the search results are poor, the conversion rate is not high. In this paper, a method based on analysis of time series is proposed to calculate the scores of partial search factors by using prediction data, which can satisfy with the proper ordering of some periodic seasonal goods in the commodity search.