可变多阶Markov链模型及在WWW个性化推荐中的应用

The Variable Multiple Order Markov Chain Model and Its Appfication in WWW Personalized Recommendation

WWW个性化推荐问题是WEB挖掘的一个重要研究方向。针对传统的固定阶数的Markov链模型用于WWW个性化推荐问题的不足，提出可变多阶Markov链模型（Variable Multiple Order Markov Chairr--VMOMC）。VMOMC将用推荐目标网页概率值度量的可变多阶Markov链并行组合，组合模型中采用遗传算法确定各单阶Markov链模型的最优权重，在真实WWW站点访问日志数据上的实验结果表明：VMOMC与传统的定阶Markov链浏览推荐模型相比，具有较理想的推荐性能。

WWW personalized recommendation is an important research direction of Web mining. Taking into consideration the problems existing in WWW personalized recommendation using a fixed order Markov chain model, the Variable Multiple Order Markov Chain Model （VMOMC）is proposed. VMOMC parallel combines variable multiple order Markov chains which assign each web page of recommendation result with a probability measurement value. Genetic algorithm is used for optimizing the weights of each single order Markov chain in the combination model. Compared with the traditional fixed order Markov recommendation model, experiment result with a real WWW Web access log data shows that VMOMC has improved the performance of recommendation.