Abstract: This paper proposes an association recommend item generating and sorting method for comprehensive e-commerce based on time weighting. Mixture of vertical, horizontal data formats and hash technology are employed to obtain the recommended candidates efficiently. Improved Jaccard similarity coefficient with time factor is taken into calculation, meanwhile it combines with the degree of association for the candidates sorting to balances the diversity of recommendation and consumers’ preference. Application results show that this method has better performance in recommend-items generating efficiency and online recommending effect.

Key words: association recommendation, time weighting, comprehensive e-commerce, dictionary hash, degree of association

摘要： 提出一种基于时间加权的综合电子商务物品关联推荐项生成和排序方法。通过垂直和水平2种数据格式以及字典散列技术，快速得到物品的候选推荐项。采用时间加权计算方法修正Jaccard相似度系数，结合关联度得到候选推荐项的排序，从而保证推荐的多样性，并兼顾用户的消费喜好。应用结果表明，该方法能提高推荐项的生成效率，具有较好的在线推荐效果。

关键词: 关联推荐, 时间加权, 综合电子商务, 字典散列, 关联度

CLC Number: 

• TP18