本文论述了Web用户访问模式挖掘中的数据预处理,主要提出了数据预处理中如何识别会话的一种改进算法。该方法通过使用三个因素来构造会话:①根据先验知识,确定会话时间阈值识别会话;②根据页面访问时间统计分布,确定相邻网页访问时间间隔阈值识别会话;③页面内容及站点结构确定页面重要程度识别会话。实验结果表明,相对于传统的单一方法进行会话识别的方法,该方法能够准确的识别会话,更为合理有效。 This paper discusses the data preprocessing in Web user access pattern mining, and proposes an improved algorithm for how to identify the session in data preprocessing. The method constructs the conversation by using three factors: ① Identify the conversation according to the prior knowledge; Identify the conversation according to the statistical distribution of the page access time; ③ Identify the conversation by the threshold of the adjacent webpage access time; ③ Identify the content of the page and the structure of the site Page importance recognition session. Experimental results show that compared with the traditional single method of session identification, this method can identify the session accurately and is more reasonable and effective.