نوع مقاله : مقاله پژوهشی

نویسندگان

• احسان الله خوشخوی نیلاش ¹
• منصور اسماعیل پور ²
• بهروز بیات ³
• علیرضا اسفندیاری مقدم ⁴
• عرفان حسن نایبی ⁵

¹ دانشجوی دکتری مدیریت فناوری اطلاعات،گروه مدیریت، واحد همدان، دانشگاه آزاد اسلامی، همدان، ایران

² دانشیار، گروه مهندسی کامپیوتر، واحد همدان، دانشگاه آزاد اسلامی، همدان، ایران  نویسنده مسئول: esmaeilpour@iauh.ac.ir

³ استادیار، گروه علم اطلاعات و دانش‌شناسی، واحد همدان، دانشگاه آزاد اسلامی، همدان، ایران

⁴ استاد گروه علم اطلاعات و دانش‌شناسی، واحد همدان، دانشگاه آزاد اسلامی، همدان، ایران

⁵ استادیار، گروه مهندسی صنایع، دانشگاه صنعتی شریف، تهران، ایران

چکیده

بانک‌ها برای تسهیلات سرمایه‌ثابت از فرایند‌های پیچیده و طولانی‌مدت شامل مراحل، نقاط کنترل و تایید زیاد برخوردارند. اینگونه فرایندها در ایجاد و توسعه واحدهای صنعتی، معدنی و گردشگری از اهمیت فزاینده‌ای برخوردارند. تجزیه و تحلیل مداوم اینگونه فرایندها برای ارتقاء، بهبود مستمر و کسب دانش از انها اهمیت فزاینده‌ای دارد. هدف اصلی تحقیق حاضر ارائه چارجوب روش‌شناختی جامع بر پایه فرایندکاوی با ترکیب با داده‌کاوی در خصوص تجزیه و تحلیل فرایندهای تسهیلات سرمایه ثابت است. روش‌شناسی بکاررفته در پژوهش حاضر برگرفته از تکنیک‌ها و مفاهیم فرایندکاوی و داده‌کاوی بر پایه داده‌های رخداد سیستم تسهیلات، بانکی فعال در ایران است. این روش شامل نه فاز شروع، آماده‌سازی، بازرسی، کاوش و تحلیل، ارزیابی، تحلیل چند بعدی فرایندی، پیش‌بینی، بررسی نتایج و بهبود است. هر کدام از فازها شامل چندین مولفه و به صورت تکرارشونده است. از جمله نتایج پژوهش حاضر استخراج فرایند واقعی، شناسایی گلوگاه‌ها، مراحل پرتکرار در نمونه و دارای فراوانی و گونه‌های فرایند است. افزون بر این شناسایی شعب و افراد دارای بیشترین نقش و ویژگی‌های داده‌ای موثر بر کاهش زمان پرداخت تسهیلات، تحلیل فرایند از ابعادی ماننداستان از دیگر یافته‌ها بود و نشان داد تجزیه و تحلیل استانی دانش خوبی در اختیار می‌گذارد. یکی از ابتکارات تحقیق حاضر استفاده از داده‌کاوی برای پیش‌بینی زمان پرداخت تسهیلات بود. در مقایسه روش‌های متنوع، الگوریتم درخت تصمیم‌گیری با دقت 72 درصد بهترین عملکرد را داشت.در نهایت علاوه بر کشف انحرافات، بر مبنای ایجاد داده‌های رخداد و تحلیل آن، فرایند بهبود‌یافته استخراج شد،که نشان از بهبود 67 درصدی بود.

کلیدواژه‌ها

• فرایندهای سرمایه ثابت
• چارچوب روش‌شناختی
• داده‌های‌رخداد
• فرایندکاوی
• داده‌کاوی

موضوعات

• مدیریت داده، اطلاعات و دانش در حوزه کسب‌وکار هوشمند

عنوان مقاله [English]

A methodological framework for the analysis of Facility processes based on process mining and data mining: a case study of the fixed capital facilities processes

نویسندگان [English]

• Ehsan allah Khoshkhoy Nilash ¹
• Mansour Esmaeilpour ²
• Behrooz Bayat ³
• Alireza Isfandyari Moghaddam ⁴
• Erfan Hassannayebi ⁵

¹ PhD Student of Information Technology Management, Department of Management, Hamedan Branch, Islamic Azad University, Hamedan, Iran

² Associate Professor, Department of Computer Engineering, Hamedan Branch, Islamic Azad University, Hamedan, Iran Corresponding Author: esmaeilpour@iauh.ac.ir

³ Assistant Professor, Department of Knowledge and Information Science, Hamedan Branch, Islamic Azad University, Hamedan, Iran

⁴ Professor, Department of Knowledge and Information Science, Hamedan Branch, Islamic Azad University, Hamedan, Iran

⁵ Assistant Professor, Department of Industrial Engineering, Sharif University of Technology, Tehran, Iran

چکیده [English]

Banks have complex and long-term processes for facilities, including many stages, control points and approvals. Continuous analysis of such processes is increasingly important for continuous improvement and gaining knowledge from them. The main goal of the present research is to provide a comprehensive methodological framework based on process mining and data mining regarding the analysis of fixed capital facility processes. The method used in the present research is derived from the techniques of process mining and data mining based on the event log of the facility system, an active bank in Iran. This method includes nine phases of initiation, preparation, inspection, exploration and analysis, evaluation, multi-dimensional analysis, prediction, review of results and improvement. Among the results of the present research is the extraction of the real process model, identification of bottlenecks, frequent activities in a case and all cases and process variant. In addition to this identification of branches and people with the most important roles and based on data features in reducing the time of payment of facilities, the analysis of the process from dimensions such as the province was one of the other findings. One of the initiatives of the present research was the use of data mining to predict the payment time of facilities. In the comparison of various methods, the decision tree algorithm had the best performance with 72% accuracy. In addition to identifying deviations, based on the creation of event log and its analysis, the improved process of extracting which showed a 67% improvement.

Introduction

Today's businesses benefit from a number of processes in order to earn more income and better services (Dakich et al., 2018). They are looking for processes that have better and more successful performance in order to achieve organizational goals and optimal use of resources in the operational environment (Van Der Aalst, 2016). Therefore, continuous analysis of processes for continuous improvement in organizations is very important.
Considering that the processes of providing facilities, especially fixed capital, are very effective in the creation and development of industrial, mineral and tourism units, having knowledge of them is of increasing importance. One of the efficient and effective methods for analyzing and improving business processes is process mining. With the help of its various concepts and techniques this method provides useful knowledge for the detailed examination of processes and how they are realized.
On the other hand, the efficient method of data mining, which provides the possibility of extracting knowledge from historical and predictive data (Basha, 2017), can be combined with the process mining method. With the investigations carried out, the methodological framework in order to provide process-centric and data-centric analysis, including the discovery of the real process model of facility payment, performance analysis of such processes, analysis of process varints, multi-dimensional process-centric analysis, payment time prediction, recommendations for improvement and process improvement based on event log simulation is not presented. Also, due to the novelty of the process mining method, the purpose of this research is to provide a comprehensive methodological framework using these techniques, concepts and tools of process mining in combination with data mining methods regarding the analysis of business processes with the study of fixed capital facilities processes.
Research Question(s)
How to provide a methodological framework for the analysis of fixed capital processes by using the techniques and concepts of process analysis and data mining methods?

Literature Review

In Table No. 1, a number of related studies are compared with each other.
Table 1.  Summary of the research conducted




Research


Business


Components used


Event log


Miners






(Urrea-Contreras et al., 2017)


SME organizations


Event Log extraction, discovery, conformance checking, extend model, and return integrated model


software development system (JIRA)


inductive




(EL KODSSI & Sbai, 2024)


Smart environments


Data selection, data transformation, generation of event log, discovery, enhancement


Unstructured sensor generated data


MDA and machine learning




(Rashed et al., 2023)


hospital


Preprocessing, model discovery and analysis


Heart surgery unit in a hospital in Egypt


heuristic, inductive, ILP and ETM




(Erdogan & Tarhan, 2022)


Emergency


Determining goals, extracting event log, pre-processing, applying multi-perspective process mining, analysis, recommendation for improvement and evaluation of results.


Emergency system log


fuzzy




(Pan & Zhang, 2021)


Construction project


Event log generation and preparation, discovery and validation


Example of a construction project


Fuzzy and inductive




(Lorenz et al., 2021)


Production business


Mapping, analysis and improvement


Production business event log


fuzzy




(Augusto et al., 202)


Healthcare trends


Planning, data extraction, data processing and evaluation


Patients in Victoria, Australia


fuzzy




(Pang et al., 2021)


Acute care and treatment processes


Coding and categorizing activities, extracting and filtering event log, discovering and improving the process model and performance analysis


Stroke care process


IDHM miner, alpha, fuzzy and heuristic




(Ramos et al., 2021)


ERP configuration, intelligent agriculture and computer configuration


Extract configuration event log, control and clean data based on feature model, build data clusters and discover related workflow.


 


Greed, hierarchy and genetics




A number of studies are not comprehensive in using the concepts of data mining and process mining. Some of them lack features such as multidimensional process centric analysis, event log simulation for improvement, evaluation of results with field specialists and so on. Comparing the studies, each of these cases can be expressed as a research gap. It is also necessary to consider all the components and phases as a methodological framework as another research gap.
 

Methodology

The method used in the present research is based on the techniques, concepts and methods of the process mining in its manifest (Will van der Alast et al., 2011). In this research, the event log of the fixed capital facility system of one of the active banks in Iran has been used. The proposed framework includes nine phases of initialization, preparation, inspection, analysis, evaluation, process centric analysis, prediction, transfer results and finally improvement. Figure 1 depicts the mentioned methodological framework.
Figure 1. The mentioned methodological framework
 

Results

Process models were discovered based on alpha, alpha++, heuristic, genetic, fuzzy and inductive techniques. By comparing inductive and fuzzy model, fuzzy model is very effective due to less edge filter and coverage of all activities. Process bottlenecks, people and branches with the most important roles were identified.
The heuristic algorithm with a value of 0.833 had the best performance in the average values of the quality indicators of the process model. In Figure 2, the mentioned methods are compared.
Figure 2. Comparison of miners
 
Analyzing the impact of data features with a target throughput time of 271 days, according to the dimensions of the Civil Partnership Bases contract, Riyal Civil Partnership Contracts and SME customers had the greatest impact in reducing the process throughput time.
The J48 decision tree algorithm had the best performance with 72% accuracy compared to all the data mining methods used.
Figure 3. Results of data mining analysis with J48 algorithm
 
203 records were used to simulate new event data. The results of the analysis showed a 67% improvement.
Keywords: Fixed capital processes, methodological framework, event log, process mining, data mining.
 
 
 
 

کلیدواژه‌ها [English]

• Fixed capital processes
• methodological framework
• event log
• process mining
• data mining