Research On Process Mining Method For Duplicate Tasks

As business processes become more and more complex,the number of events logs recorded by business processes is increasing.Process mining aims to extract knowledge from event logs,which includes process discovery,conformance checking and process enhancement.Process discovery aims to mine the model from the event logs(the mined model),conformance checking aims to evaluate the quality of the mined model(i.e.,the performance of process discovery methods),and process enhancement aims to repair or improve the mined model.One of the difficulties in process mining is that the mining process models contain duplicate tasks.And most modern conformance checking methods cannot effectively evaluate the quality of the mined model containing duplicate tasks.The contributions to solve these problems are summarized as follows:1.The thesis investigates and analyzes the the current process discovery methods that can discover duplicate tasks,and the conformance checking methods that can evaluate duplicate tasks,then finds the problem that the different process discovery methods chose their own datasets or evaluation metrics.To solve this problem,the thesis proposed a process mining method for duplicate tasks,which designs selection guidelines for process discovery methods,evaluation metrics and datasets.The methods can find the best process discovery method to discover duplicate tasks in different quality dimensions.Finally,the thesis conducts an experiment to apply the guidelines.2.The exsiting simplicity evaluation methods cannot distinguish between duplicate tasks and invisible tasks,fail to consider the model size and the basic assumptions in these methods are not consistent.To solve this problem,the thesis proposed a simplicity evaluation method by combining the types of tasks.The method counts the number of differents tasks in the original model and mined model and their weight values.Then the simplicity is calculated by comparing the number of different types of tasks in the mined model with the original model.The case analysis and experiment results demonstrate that the proposed method can effectivelyevaluate the simplicity of the mined models containing duplicate tasks.3.The existing structuredness evaluation methods cannot evaluate the distribution of duplicate tasks in different process structures.To solve this problem,the thesis proposed a structuredness evaluation method for duplicate tasks.The method divides process structures into different components which can be used to identify and count duplicate tasks.And then uses the confusion matrix is used to count the number of duplicate tasks in different process structures from the originaland mined model.The accuracy of the model to discover duplicate tasks in different process structures is calculated according to the statistical results.The case analysis and experiment results demonstrate that the proposed method can effectively evaluate the accuracy of process discovery methods to discover duplicate tasks in different process structures.