基于网络日志的用户行为刻画与预测研究

Analysis and Prediction of User Behavior Based on Web Log

采用基于相似度的特征聚类算法以及粗糙集模糊分析法,提出了基于网络日志的用户性格特征分析及行为预测方法.首先,构建标准性格特征向量库;然后,采用基于余弦相似度的特征聚类算法进行性格分析,该算法解决了适量样本情况下的机器学习中聚类的问题,使训练模板数据即使在数据不是足够大的情况下仍能提取特征;最后,采用基于粗糙集理论的模糊分析算法进行行为预测,该分析算法简化了分析过程,减少了建模中需考虑的因素,又能得出精确的结果.对比实验表明,该方法能较准确地分析不同用户性格特征和对其未来行为进行预判,并分析出可能对安全领域造成威胁的人群.

The feature clustering algorithm based on similarity and the fuzzy analysis method based on rough set were used.A method of analysis and prediction of user behavior based on web log was proposed.Firstly, a standard character eigenvector library was constructed.Then, a character clustering algorithm based on cosine similarity was used for character analysis.Finally, a fuzzy analysis algorithm based on rough set theory was used to perform behavior prediction.Results showed that the method accurately analyze the personality characteristics of users and predict their future behaviors, and identify the groups that might pose a threat to the security field.