Application of Clustering Filters in Order to Exclude Irrelevant Instances of the Process Before Using Reinforcement Learning to Optimize Business Processes in the Bank

Abstract

The research offers and describes the use of clustering filters in order to exclude preliminarily the instances of the process, which contain errors, and which are not related directly with the business process, and, accordingly, are irrelevant for the analysis. Comparison of 15 types of filters was performed using mapped-out data. It was shown that successful preliminary filtering is possible before the application of reinforcement learning for business process analysis, which reduces the data processing amount.