基于日志活动捆绑挖掘隐变迁

Mining Hidden Transition Based on Log Activity Binding

研究隐变迁挖掘问题,提出了过滤低频日志的预处理方法。其中,使用编辑距离函数计算日志序列与初始模型行为子集的总体近似适合度及其上下界;同时,放宽适合度的区间定义,仅去除最可能是噪声的日志,而留下符合正常流程的低频日志。此外,提出了日志活动捆绑的定义,为找到有效低频序列中每个活动的直接输入和输出捆绑活动集,进而挖掘出模型中合理的隐变迁提供理论支撑。

A preprocessing method is proposed for filtering low frequency log to study the mining hidden transitions.The overall approximate fitness and upper-lower boundary of the log sequence and the subset of the initial model behavior is calculated through the edit distance function.At the same time,the interval definition of fitness is relaxed to remove the probable log noise,leaving a low frequency log meeting the normal process.In addition,the definition of binding among activities in the log is proposed to find the direct input and output binding activity set of each activity in the effective low-frequency sequence,and then the reasonable hidden changes are excavated in the model.