Task Clustering Method Using User Interaction Logs to Plan RPA Introduction

Abstract

Robotic process automation (RPA) software is a powerful tool that can automate business operations to reduce manual labor while improving operational quality by eliminating input errors. In order to efficiently and effectively improve business operations with RPA, it is necessary to clarify the types and volumes of actual business operations being performed by the employees and improve operations that have a large volume and are performed repeatedly. User interaction (UI) logs consist of users’ activities performed on the computer and can be collected regardless of the business system or application to understand how employees work. However, it is difficult to understand the types and volumes of the executed tasks from such data because the task types are not recorded explicitly. In this work, we propose a method that clusters UI logs into task types to help analyzers identify high-volume and repetitive tasks for RPA introduction. As the operation types differ by task type, we utilize this characteristic to analyze the co-occurrence of operations and segment UI logs into a sequence of the same task types. Then, we perform clustering based on the operation types contained in the segments. We evaluated our approach using UI logs generated from actual scenarios in a workplace, and report the results and limitations.