Repetitive task process discovery method based on equivalence class

Abstract

The present invention discloses a repetitive task process discovery method based on an equivalence class. The method comprises: by extending an event order relation definition, proposing a determination rule for the equivalence class to divide a same task subset; in a pre-processing stage, dividing events with the same names with a correct dependence order relation into a same task equivalence class subset; in a processing stage, discovering an algorithm to generate a WF-net (Workflow Petri net) model by using a short cycling process; and in a post-processing stage, renaming different same task subsets to eliminate repetitive tasks. According to the method, a standard service process XES file is first read, the dependence order relation definition and the determination rule are adopted to detect the repetitive tasks, and a service process model represented by adopting WF-net is output. The method has relatively good capability to solve the problem in discovery of multi-predecessor-successor repetitive task process with the short cycling structure, and can ensure a correct and reasonable mining result.

Description

technical field [0001] The invention relates to the field of data processing, more specifically, to a method for discovering repetitive task processes based on equivalence classes. Background technique [0002] At present, various information systems based on distributed computing, Internet of Things, and service computing environments all have a large amount of log data that records the process of business execution. Process mining technology analyzes these logs to discover valuable knowledge and help improve the original business process. Process discovery is the core technology of process mining, that is, how to discover business process models from logs. Usually business models include control flow structures such as sequential, parallel, selective, cyclic, non-free choice, implicit tasks, and repeated tasks. The ability to mine control flow structure, that is, to construct a process model that can describe the relationship between activity execution from event logs, i...

AI Technical Summary

Problems solved by technology

[0006] The above-mentioned method 1 cannot solve the problem of repetitive task recognition that contains short loop structures, because the heuristic rules of machine learning are not capable enough, and because no formal theory is used to represent the generative model, the mining results may be unreasonable; method 2 expands the heuristic rules to identify non-free choice structures, but still does not solve the shortcomings of method 1; methods 3 and 4 can only deal with short loops and repetitive task discovery problems in some cases, because they do not consider the situation where repeated tasks appear in different execution trajectories, and there is no Handle siege cases with multiple predecessors and successors

Images