Log-Driven Behavior Discrimination of Process Robots Using Entropy Similarity

In the context of Robotic Process Automation, process robot refers to robot software corresponding to a specific process, with the ability to capture and interpret process management. It is a fact that many process robots are derived from the same base model in practical applications, and the main r...

Full description

Saved in:
Bibliographic Details
Main Authors: Lu Li, Huan Fang
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10870220/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the context of Robotic Process Automation, process robot refers to robot software corresponding to a specific process, with the ability to capture and interpret process management. It is a fact that many process robots are derived from the same base model in practical applications, and the main reasons include the inevitable and flexible adaptability of process models. This fact raises the question of many process robots having a high degree of similarity, and it is hard to differentiate them from each other under the state-of-the-art similarity measurements. In order to differentiate similar but different process robots, in this paper we propose an approach to process robot behavior discrimination, from the perspective that only event logs are given as known, that is, the reference models of process models are kept unknown. The biggest innovation lies in that the proposed behavior discrimination method is based on log-entropy similarity measurement using probability distributions and statistical distance theory. Specifically, using the concept of log entropy, event logs are transformed to probability distributions of trace multisets and activity pairs multisets, respectively. Then, JS and KL divergence is used to measure the statistical distance between the two probability distributions of processes, and these statistical distances are the foundations of log similarity measurements. Finally, a series of experiments are conducted on a set of publicly available datasets, validating the feasibility, generality, and effectiveness of the proposed method.
ISSN:2169-3536