As organizations increasingly work in process-oriented manner, the number of business process models that they develop and have to maintain increases. As a consequence, it has become common for organizations to have collections of hundreds or even thousands of business process models. When a collection contains such a large number of business process models, it is impossible to manage that collection manually. Therefore, Business Process (BP) Model Repositories are required that store large collections of process models and provide techniques for managing these collections automatically and efficiently. The goal of research described in this thesis is to improve on existing BP Model Repositories, by improving the management techniques that are supported by these repositories on an aspect that has received little attention so far. Looking ahead at the results of the research, the aspect that will be selected for improvement is the process retrieval aspect. The two main research activities that will be carried in the context of this research are the following. Firstly, a survey of Business Process Model Repositories is performed to identity an unsolved aspect to be enhanced. The functionality of existing BP Model Repositories is listed and summarized as a framework for BP Model Repositories. After comparing the functionality that is provided by existing BP Model Repositories, based on the framework, efficient process retrieval is selected as the aspect that will be improved. This aspect is selected, because, although existing BP Model Repositories provide techniques for process retrieval, none of them focus on the efficiency of process retrieval. Secondly, an indexing technique for process retrieval (both process similarity search and process querying) is proposed. The index is constructed using features of process models. Features are small and characteristic fragments of process models. As such, by matching features of a given query/search model and features of a process model in a collection, a small set of models in the collection that potentially match the query/search model can be retrieved efficiently through the index. Techniques are also proposed to further check whether a potential match is an actual match for the query/search model. All of the above techniques are implemented as a component of the AProMoRe (an Advanced Process Model Repository) process repository. To evaluate the proposed process retrieval techniques, experiments are run using both real-life and synthetic process model collections. Experimental results show that on average the process retrieval techniques proposed in this thesis performs at least one order of magnitude faster than existing techniques.
|Qualification||Doctor of Philosophy|
|Award date||25 Sep 2012|
|Place of Publication||Eindhoven|
|Publication status||Published - 2012|