Abstract: A technology based on maximal frequent sequential pattern direct graph is presented to perform personalization recommendation.Because of taking the access sequence of pageviews into account in this technology，a higher accuracy can be obtained.Direct graph structure compresses and stores all maximal frequent sequences discovered from Web usage mining，recommendation engine matches some sub paths of the direct graph and has a recomentdation both on breadthwise and lengthways according to nearest access subsequence form user active session，without need to find the identical or similar patterns in the whole pattern database，which enormously improves the speed of patterns matching to satisfy the need and specific of real-time recommendation better.The results of experiments show that the approach is effective in speed and precision.

Key words: maximal frequent sequential pattern, personalization recommendation, breadthwise recommendation, lengthways recommendation

摘要： 提出一种基于最大频繁序列模式有向图的页面个性化推荐技术，由于考虑了用户会话的页面访问顺序，比一些不考虑页面访问顺序的推荐技术有更高的准确率。有向图结构压缩存储了所有最大频繁序列模式，推荐引擎依据截取的用户最近访问页面子序列，与有向图的部分路径进行匹配并进行横向推荐和纵向推荐，无需在整个模式库中搜索相同或相似的模式，从而加快了模式匹配的速度，更好地满足了页面推荐的特性和实时要求。实验证明，方法是有效的。

关键词: 最大频繁序列模式, 个性化推荐, 横向推荐, 纵向推荐