Authors:
Shinji Kikuchi
1
;
Yasuhide Matsumoto
1
;
Motomitsu Adachi
1
and
Shingo Moritomo
2
Affiliations:
1
Fujitsu Laboratories Limited, Japan
;
2
Fujitsu Limited, Japan
Keyword(s):
Process mining, Batch job, Job net, Integrated system, Behavior analysis.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Business Analytics
;
Business Process Management
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
e-Business
;
Enterprise Engineering
;
Enterprise Information Systems
;
Health Information Systems
;
Information Engineering Methodologies
;
Information Systems Analysis and Specification
;
Knowledge Management and Information Sharing
;
Knowledge-Based Systems
;
Methodologies, Processes and Platforms
;
Model-Driven Software Development
;
Modeling of Distributed Systems
;
Sensor Networks
;
Signal Processing
;
Soft Computing
;
Software Engineering
;
Symbolic Systems
;
Systems Engineering
Abstract:
Batch jobs, such as shell scripts, programs and command lines, are used to process large amounts of data in large scale enterprise systems, such as supply chain management (SCM) systems. These batch jobs are connected and cascaded via certain signals or files so as to process various kinds of data in the proper order. Such connected batch jobs are called “job nets”. In many cases, it is difficult to understand the execution order of batch jobs in a job net because of the complexity of their relationships or because of lack of information. However, without understanding the behavior of batch jobs, we cannot achieve reliable system management. In this paper, we propose a method to derive a job net model representing the execution order of the job net from its logs (execution results) by using a process mining technique. Improving on the Heuristic Miner algorithm, we developed an analysis method which takes into account the concurrency of batch job executions in large scale systems. We
evaluated our analysis method by a conformance check method using actual job net logs obtained from a large scale SCM system. The results show that our approach can accurately and appropriately estimate the execution order of jobs in a job net.
(More)