Combi-BP: automating the data-oriented optimization in business processes. From declarative to executable models.

Abstract

One of the main objectives of a business expert is to model the business goals of an enterprise process. Several languages have been created to describe the necessary activities to achieve the objective, especially in the business process context. These languages can be divided into imperative and declarative ones. Declarative languages tend to be used when the speci c model is unknown, being possible to describe what has to be done instead of how. Otherwise, imperative languages permit to describe how the things have to be done and then, the imperative models can be executed in any Business Process Management System (BPMS). The declarative descriptions are more exible, since they permit to describe the model in a more relaxed way, which means that various process executions can follow the same declarative description. However, both paradigms are focused on the activities order description, but unfortunately, the data perspective is missed. Furthermore, the optimization of a business goal which depends on the exchanged data during the execution of the business process has not been included in previous proposals. There are no solutions that allow the business experts to describe nor execute a declarative description where the executed model depends on the exchanged data between the involved activities in each instance. In this thesis dissertation, an approach to support this data-oriented optimization in business process is presented. A data-oriented optimization problem is a process whose main purpose is to obtain the best business product. In order to obtain this business product, the process must combine several activities by taking into account the existing data-structure and data-value dependencies. Both kind of dependencies are established by a set of constraints that relate the data (consumed and provided by the activities) and the data given by the customer. Therefore, the BPs under the scope of our research are those which are centred on developing sound data in business processes, analysing how data-structure and data-value dependencies can aect the correct business process execution. However, if the data provided at runtime for the activities that conform the model have not got enough level of quality, then business process will not be successfully executed. The base of the proposal is focused on the combination of the advantages of both paradigms: the exibility of the declaratives, and the automatic execution in a BPMS of the imperatives. On the one hand, we want to describe a exible model using a declarative description where the exchanged data and an optimization objective are included. In the other hand, this declarative model must be executed in a generic business process management system with the aim of support any instance of the process. Therefore, how the declarative description can be transformed into an imperative business process is developed. The transformation methodology that we propose is based on Model-Driven Architecture. Firstly, the declarative is transformed into an imperative which takes into account the data-structure dependencies. The exibility of the declarative speci cation is kept thanks to the use of Constraint Programming. On the other hand, the resulting imperative model is enriched with new intelligent techniques, also based on Constraint Programming, in order to solve the data-value dependencies. Finally, a methodology and an implementation are developed in order to make the business process aware of the data-quality aspects.