A Algorithm For Detecting Concept Drift Based On Context In Process Mining

Process mining plays a very important role in our daily production work. However, due to the competition in the market and technology upgrades, process will tend to be flexible and Elasticity. We need to perceive the impact of process system’s changes and changes to the current system operation. It is important to optimize and control changes in the allocation of material and human resources at the same time. In the complex process, it is very difficuity to check the process’s tiny change point by people. The enterprises need a kind of technology to detect the process’s change quickly and automaticly. So enterprises could analyse the problem and solve the problem. We call this technology, in the field of process mining, concept drift detection based on process mining.The process context application will be the next breakthrough in the field of process mining. In this paper, we araise a new kind of algorithm based on the original study. The algorithm detect process concept drift by calculating the structural properties of the sample (correlation count, the relationship entropy, following matrix), as well as the context attributes (time attributes, personnel attributes) and using the attribute matrix to calculate the distance between the logs, at the same time, depending on hypothesis testing technology. In order to lower the time complexity, the algorithm uses the stability of the process before and after the change points. For the purpose of improving the sensitivity of the algorithm to detect process changes, the algorithm uses the context of the time change and worker change. Finally, we design related experiments to verify the algorithm. We find that when the log scale in200and2000case, the proposed algorithm can find all the drift, and does not appear the mistake porint. But the original algorithm in the200and2000log scale was appear loss the change point ore detect the mistake point.Through the discussion and experimental, we can see that the proposed algorithm, both in the theoretical design and experimental results with respect to the existing research results have progress from the time efficiency and the accuracy.