Research On Alignment Of Event Logs And Model Behavior And Process Model Solf-adaptation

Various Process-Aware Information Systems have recorded a large amount of data.And these data form the foundation of process mining.Process mining relates the recorded event log with the process model,and it can find the problems in event logs or process models by comparing of their behaviors.There are some problems in current research of process mining:(1)alignment of event logs and process model:the alignment algorithms for declarative process model and event logs are not efficient;the alignment methods for large-scale complex models have some limitations;the existing alignment methods involve only activities' events attributes,and other attributes are often ignored;(2)due to changing requirements,the researches about process model self-adaptation should be further studied.On account of these problems,this dissertation studies the process model alignment and the adaptability of the process models in the base of Petri nets and its constraint forms.The main work and innovations are as follows:(1)Propose an alignment method based on declarative process tree modelA new alignment method is proposed to align the declarative process model represented by a process tree and its event logs.An eliminating backtracking method is presented for the possible backtracking in alignment,and the FirAct and FolAct set of each sub-tree are limited to avoid potential undetermined move matching.Then,the alignment conditions of the process model and the alignment algorithm are given.Taking the emergency clinical path of a hospital in Qingdao as an example,the algorithm is compared with the existing alignment algorithm.Experimental results show that the proposed algorithm is superior to the existing alignment algorithm.(2)An alignment algorithm based on sub-alignment selection and merging rulesThe merge conflicts and pseudo-alignments problem are discussed in alignment based on process model decomposition.Selection and merging rules of sub-alignments are proposed to improve the quality of merg,ing alignment and lower the pseudo-alignment ratio.Meanwhile,the relationships are discussed among sub-alignments,monolithic alignment and strong similarity alignment.And then,an alignment method is proposed to enhance the diversity of merged alignments.Taking the clinical pathway of stenocardia in a hospital in Qingdao as an example,the comparison among the proposed algorithm and the existing algorithm are carried out.The experimental results show that the algorithm is superior to the existing alignment algorithm from the quality of the merged alignment to the efficiency of the merged alignment.(3)Multi-perspective alignment method based on event similarityThe alignment method is discussed based on multiple attributes of events.Firstly,the event attributes can be annotated in the functional domain ontology tree,and the similarity of the event attributes is obtained based on the distance of the event attributes in the ontology tree.And then,the similarity of events is determined according to attributes similarity.A multi-perspective alignment algorithm based on event similarity is presented.The relationships between multi-perspective alignment and simple alignment are analyzed.And experimental results show that the proposed algorithm is superior to existing simple alignment and multi-perspective alingment methods in fitness and precision.(4)Model self-adaption method based on process variant commonnessBased on hiding operation,blocking operations and inheritance relationship among process models,an algorithm is presented to extract process variant commonalities-core process fragments.Then,the existing process model can be extended based on the variation commonness of process model.The extended model can preserve the characteristics of original model while embodying the commonness of process variation,so it has better environmental adaptability.Finally,an example is given to illustrate the algorithm.