With the rise of globalisation businesses have to face omnipresent competition in a fast-changing environment characterised by increasing complexity. As a result, the long-term existence of enterprises depends on the ability to react to internal or external business stimuli in a fast and flexible manner. The early identification and effective compensation of environmental changes in terms of their utilization as competitive advantages is commonly accepted as a crucial strategic success factor. This circumstance necessitates among other things the agile development of new forms of businesses that cross organizational borders as well as the flexible optimization and adaptation of existing business processes. Therefore, this section discusses the current state of the art in the field of process optimization especially on the base of mining technologies as well as the latest developments in the field of process adaptation.
Business process optimisation. In the field ofbusiness process optimisation the work of [Vergidis 2007] is a central contribution. This work builds upon the mathematical model of [Hofacker 2001] utilising multi–objective optimisation algorithms instead of a single objective GA approach. [Vergidis 2008] puts forward a multi-objective optimisation framework, based on an evolutionary technique, to allow for a mined process to be optimised in terms of its task attributes such as duration and cost. In later work [Vergidis 2008] utilises web services as the task elements of a process allowing for a library of suitable services to be built and used in the optimisation of processes constructed by a genetic algorithm. At present this work does not optimise processes in real-time and the re-configuration of processes is limited to the random substitution of web service based tasks from a pre-defined library. An initial experimentation in the expansion of the web service library (increasing the amount of web service categories and tasks) has been completed by [Tiwari 2011], and shows promising results. The business process optimisation research outlined in this proposal is different from that concentrating on web service composition [Agarwal 2008] in that the process itself is being built based on live data with only limited guidance on what exact role the process will perform. The practice of redesigning a process is an active research topic as relevant literature still restricts itself to the description of ‘situation before’ and ‘situation after’, giving little information on the re-design process itself [Reijers 2005]. At present, therefore the majority of manual and automated attempts to improve business processes concentrate on simple sequential processes.
Process adaptation. To overcome these obstacles, in the last decade, considerable research effort was spent on systems for supporting flexible processes throughprocess adaptation. In dynamic environments where business requirements continuously change, processes need to be able to adapt to fluctuating constraints. Processes cannot remain rigidly structured but need to support ad-hoc human control and unforeseen alternative execution paths.Depending on the supported level of flexibility, we can distinguish between roughly three types of processes:
A study by [Burkhart 2010] highlights that while ad-hoc and case-based process systems provide the required flexibility they lack support to automatically adapt a process and recommend new courses of action to the user. Alternatively, systems for adaptive structured processes can react to certain context changes automatically, but do not offer the flexibility for the user to control the process flow. To cope with this drawback, a wide range of differing approaches, concerning process flexibility during runtime, have been proposed by research and development; these have not so far achieved a wide-ranging acceptance or a high market penetration. This may be traced back to the fact that flexible POIS basically suffer from the trade-off between flexibility and an appropriate process-oriented support of the end-user. In fact the higher the degree of flexibility the more the user of the system is required to have profound knowledge about the process based structures of the enterprise, whereas the introduction of additional predefined structures or guidelines limits the users’ freedom of choice. Since it is a major challenge to find an appropriate balance between both requirements, flexibility and process-oriented support during runtime, there is a need for respective concepts for improving the guidance of users when executing business processes in flexible environments. A common approach to this problem is to provide deviations from a given process structure, that have been performed in former process instances similar to the active one, as templates during runtime [Schonenberg 2008]. Even though such system functionalities advance organizations in their capabilities to draw on using up-to-date process knowledge and various best-practices by giving a user the opportunity to reuse available knowledge about possible deviations, there is still a lack of process-oriented guidance [Burkhart 2011].
In this context, a survey of [Burkhart 2012] analysed the actual state-of-the-art of process guidance techniques in the field of flexible processes. It could be observed that the necessity of assisting users has only been recognized within some research prototypes. In particular, only a few and mainly very flexible systems apply such recommender mechanisms at all. Therefore, further research activities are required in order to develop novel concepts that incorporate intelligent user assistance in terms of recommender capabilities. In doing so, first and foremost profound methods are needed to gather, maintain and evolve the recommendation basis. Moreover, mature derivation procedures for the suggestion development and comprehensive functionalities for administrative controllability are required.
Lately, in area of business process adaptation, a new research direction emerged: the process instance context adaptation. For business process context adaptation, models based on defining a specific behaviour in a certain situation using a set of context adapting rules, each rule consisting of a context condition and an associated action are proposed [Bernal 2010]. The run-time context adaptivity of a business process is activated when ad-hoc changes take place in real-time due to the changes in its execution environment, without relying on any predefined adaptation rules and strategies ([Bucchiarone 2011], [Leoni 2009], [Mounira 2010] and [Adams 2005]). Hereby, the context-aware business process adaptation is achieved without interrupting the process from its normal execution. In [Xia 2008], authors propose a business process instance adaptation method which analyses and reasons upon the current process context to dynamically and efficiently re-configure the structure of the instance such that the business goals and non-functional requirements are met. The approach addresses changes that might occur in the business process operation related to business process environment (e.g. replace a business partner, resource or component change), and to the computing level context (e.g. exception handling). Business process adaptivity is in essence an optimization problem which can be approached using bio-inspired meta-heuristics. The advantage of using bio-inspired meta-heuristics is that, by finding a proper representation model and a proper fitness function, adaptivity can be easily achieved with low processing overhead. In the context of process self-adaptivity using bio-inspired meta-heuristics, the research literature reports some approaches but limited only to supply chain management problems. As with the mining of sensor data little attempt has been made to optimise sensor output in the form of a business process.
CREMA will support the optimization of service-based process models at design time and runtime. Given annotated process models are implemented, if possible, with optimally selected and composed semantic services that are available and leasable in the cloud (functional optimization). In addition, dynamic multi-objective constraint optimisation problems for process models are solved (non-functional optimization). The process optimization component offers high-precision semantic service selection and composition, and fast DCOP solving for process models and instances in the CREMA use case domains.