摘要
用户的行为序列具有的连续性特征能够很好地反映用户的购买习惯或者用户的购物偏好,而这种特性在利用显式反馈算法进行商品推荐时往往会被忽略。因此,根据用户的购物行为等隐式反馈信息,提出一种基于变长马尔科夫模型的"自适应两阶段过滤"推荐算法。该算法主要是利用概率后缀树构建变长马尔科夫模型,然后利用该模型对用户行为序列数据集进行多次分类过滤,以判断出用户对商品的购买倾向,生成用户的商品推荐列表。实验表明所提出的推荐策略具有不错的推荐效果。
User behavior sequence with the continuity features can be a very good response to user's buying habits or the user's shopping preferences, and this characteristic in the explicit feedback algorithm is recommended products tend to be ignored. Therefore, according to the user's shopping behavior and implicit feedback information, proposes a model based on variable length Markov "two stage adaptive filtering" recommendation algorithm. The algorithm is mainly using probabilistic suffix tree construction of variable length Markov model, and then uses the model of user behavior sequence data set several times of classification and filtering, to determine the user of the commodity purchase intention, user generated commodity recommendation list. Experiments show that the proposed recommended strategy has good recommendation effect.
出处
《现代计算机》
2016年第14期8-14,共7页
Modern Computer
关键词
用户购物行为
概率后缀树
变长马尔科夫
个性化推荐
User Shopping Behavior
Probabilistic Suffix Tree
Variable Length Markov
Personalized Recommendation