基于交叉熵的用户行为异常检测

Anomaly Detection of User Behaviors Based on Cross-entropy

介绍了一种基于交叉熵模型的用户行为异常检测的方法,在实验条件完全相同的情况下,通过改变模型的参数,得到了大量的实验数据。实验结果表明,与隐马尔可夫模型相比,交叉熵模型较简单、识别性能较高、特别是训练时间可以忽略不计。

A method about anomaly detection of user behaviours based on cross-entropy is introduced.By changing the model parameters,it is found that in the same experiment conditions,the distinguishing effect of CE model is superior to that of HMM,with the former being very simple and the training time negligible.