Recommendation method RMV for partner and service selection in virtual organization breeding environments based on process mining techniques

TytułRecommendation method RMV for partner and service selection in virtual organization breeding environments based on process mining techniques
Publication TypeBook
Year of Publication2014
AuthorsPaszkiewicz, Z.
Secondary TitlePhD thesis
Number of Pages168
Date Published07/2014
CityGdańsk
Słowa kluczowecollaborative processes, process mining, recommendation system, virtual organizations
Abstract

In complex, uncertain, dynamic environment, complex production and service provision at the global scale require a large set of resources and competences that one enterprise is usually not able to provide. Thus, modern provision of services and delivery of products require integration and collaboration of many diversified, specialized, autonomous units offering access to complementary set of resources and competences.

As a generic organizational structure supporting collaboration of divers units, the concept of Virtual Organizations (VO) has been coined. VO is “a network consisting of a variety of actors, called VO members that are largely autonomous, geographically distributed, and heterogeneous in terms of their operating environment, culture, social capital and goals, which conduct processes including at least one VO collaborative process in order to carry out a particular venture due to the demand from VO clients” (Camarinha-Matos, et al., 2008). VO permits to deal with complexity, pursuit for agility, and takes advantages of broad use of information technologies in economic and managerial operations. Partners collaborating within a VO are organizations - enterprises, public administration units, and non-government organizations - people, and information systems.

The success of a VO strongly depends on ability of all participating actors to efficiently and seamlessly collaborate. Good level of collaboration is achieved by appropriate selection of services and collaborators. Due to importance and complexity of partner and service selection problem, a number of computer and organizational methods has already been proposed including the concept of Virtual Organization Breeding Environment (VOBE). A VOBE is “an association of organizations with the main goal of increasing preparedness of its members towards collaboration in potential virtual organizations” (Camarinha-Matos, et al., 2008). VOBE allows potential collaborators to prepare their future collaboration with other VOBE members before a business opportunity occurs. A VOBE that consequently applies the Service Oriented Architecture (OASIS Technical Committee, 2006) is referred to as Service-Oriented Virtual Organization Breeding Environment (SOVOBE) (Picard, et al., 2010). In SOVOBE, interaction among all the actors, virtual organizations and SOVOBE infrastructure is performed with the use of services.

Two main features of each VO collaborative process are: unpredictability and emergence. The unpredictability aspect of VO collaborative processes refers to the difficulty to plan in advance the further execution of a VO collaborative process. The emergence aspect of VO collaborative processes refers to the influence of VO collaborative process instance execution on itself, i.e., decisions made during VO collaborative process instance execution impact the next activities. As a consequence, VO collaborative processes are highly unstructured.

Proliferation of information technologies and ubiquitous access to the internet via fixed and mobile devices are followed by increased number of processes that are performed by the use of electronic means. The concept of Process-Aware Information Systems (PAISs) as “a software system that manages and executes operational processes involving people, applications, and/or information sources on the basis of process models” (Dumas, et al., 2005).

Modern PAISs log enormous amount of data providing detailed information about activities and processes that have been executed. Such data are referred to as events. Event logs provide valuable insight into process instance executions. Analysis of event logs may permit to discover factors impacting efficient execution of process instances. Information about those factors may be used to improve efficiency of future process instance executions. Efficient use of information from event logs relies on ability to analyze PAISs’ data and draw business-level conclusions concerning process execution success factors.

Discovery of knowledge from large amounts of data is the domain of data mining and machine learning techniques. To capture the notion of process in data mining, the term process mining has been coined. Process mining is “a set of techniques, tools, and methods to discover, monitor and improve real business processes by extracting knowledge from event data available in today's information systems” (Aalst, 2011).

The main idea of the Recommendation Method for Virtual Organizations (the RMV method), proposed in this dissertation, is automatic discovery of activity patterns and ad-hoc generation of recommendations for VO collaborative process instances performed within a SOVOBE. An activity pattern is a set of partially ordered activities performed by collaborators that frequently occurs in many instances of VO collaborative processes.

Four ideas are the basis of the RMV method. First, VO collaborative event logs contain information about interactions among collaborators that appear during executions of various VO collaborative process instances. Second, contexts influence behavior of collaborators. Third, frequently repeatable collaborators’ behavioral patterns, called activity patterns, can be discovered through analysis of data stored in the VO collaborative events logs. Forth, discovered activity patterns can be evaluated as good or bad practices and then used in other instances of VO collaborative process instances to improve their efficiency.

The RMV method consists of two main phases: (1) Identification of activity patterns and their contexts; (2) Recommendation formulation. In the first phase, a set of activity patterns is identified. Each activity pattern contains information about its contexts, partially ordered set of activities ordered according to temporal dependencies, involved set of actors, and relations among actors. The second phase is performed on request. In the second phase, activity patterns suited to a particular context of a running VO collaborative process are selected and recommended for inclusion in further execution of the VO collaborative process instance. Once an activity pattern is selected, it is instantiated before incorporation into VO collaborative process execution. Instantiation is based on information stored in the activity pattern regarding actors and information concerning SOVOBE members provided by SOVOBE services. Selection of the best matching activity pattern from the set of recommended activity patterns and its instantiation is performed in a collaborative way by a group of collaborators. Recommendations generated by the RMV method are used by PAIS to guide user actions. As a consequence, PAIS provides a support for both flexible definition of process models and user guidance, where the guidance is based on discovered, real and actual activity patterns.

The main achievements of this dissertation are the following: (1) Identification and evaluation of existing partner and service selection methods in the area of collaborative networked organizations and service-oriented architecture in terms of application to VO collaborative process instantiation; (2) Identification and evaluation of existing activity recommendation methods in the fields of process-aware information systems, context-aware recommender systems and process mining; (3) Formal definition of VO collaborative process, activity pattern, activity pattern context and collaborative process event log; (4) Development of the activity pattern discovery and identification method that permits extraction of activity patterns and their contexts from a collaborative event log maintained by a process-aware information system; (5) Development of the method of formulation of recommendations of activity patterns for VO collaborative process executions, where a recommendation is based on the current context of VO collaborative process and activity pattern contexts; (6) Development of the activity pattern instantiation method that permits selection of missing actors and service interfaces for activity patterns discovered on the abstract or prototype level; the selection of partners and services is performed within SOVOBE constantly throughout the VO lifecycle; (7) Implementation of a prototype of the RMV method composed of the Operational Support Service, Recommendation Manager, Recommendation Monitor, MatchMaker Module, Process Miner Module, Event log Module and Operational Support for Clients Module; (8) Example integration of the prototype of the RMV method with the process-aware information system named ErGo used to support collaboration in the construction sector; (9) Example application of the RMV method to analysis of event log data from a production company, leading to non-trivial, valuable recommendations.

The RMV method is characterized by two important features: extendibility and independence. Its extendibility relies on flexible definition of a set of attributes and functions used during activity pattern discovery and recommendation. Different sets of attributes useful in a particular domain or application can be used to describe service entities, service relations, and VO collaborative process instances event contexts. Sophistication of the functions is up to the RMV method user. Such an approach permits both rough and very refined analysis of event logs. The RMV method is independent of a particular type of VO collaborative process or type of process-aware information systems. The RMV method can be applied to analysis of any event log that follows the characteristics of collaborative event log. Such independence makes the RMV method applicable to different collaborative process requirements and different business environments.

URLhttps://www.researchgate.net/publication/265125036_Recommendation_Method_RMV_for_Partner_and_Service_Selection_in_Virtual_Organization_Breeding_Environments_Based_on_Process_Mining_Techniques