Abstarct: Nowadays, workflows are used extensively in scientific experiments and organizational processes. Workflows have quickly found their way into complex industrial processes. Recently, workflow improvement has become a new emerging problem. Regardless of the dynamics and workflow changes, some of the recent workflow improvement methods are baxsed on static workflow simplification; recent methods do not have the flexibility to predict the simplified version of the workflows. Workflow changes should be taken into consideration in measurement, and the decision about workflow simplification should be baxsed on measurement updates from workflows. In the proposed method, workflow simplification is done using statistical analysis of workflow diagrams and event logs. For this purpose, first the most frequent workflow tags are extracted. This is done using similarity evaluation. The proposed method uses these tags to identify the application domain of workflow by training a learning model. Thus, the learning model can cluster workflows. For each cluster, one workflow is selected as a cluster candidate. In this workflow candidate, the most frequent tags have a higher probability to be represented. The results of the proposed method show that the use of statistical analysis is successful in estimating simpler workflow paths. In the proposed method the value of the workflow connection parameter is reduced to 0.18. This has reduced the conflict in the workflow by approximately 30%. The advantage of the proposed method is that this method can make acceptable decisions about simplifying and improving workflows, due to the fact that it does not rely on a certain range of workflows. In the construction of a databaxse baxsed on the evaluation of workflow similarities, it is possible to dynamically add data from this new workflow and predict new paths to the databaxse. Hence, the automatic workflow simplification becomes possible. Whenever we encounter a new workflow that has not already been shown to be the same, we assign it to the cluster candidate most closely resembling it. Therefore, it is possible to predict the simplifying workflow paths that we have not seen so far. Estimating the data, on average up to 87%, leads to an accurate estimation of the domain of workflow.