On optimising resource planning for business processes

Business process management (BPM) systems are IT support tools which aim to support the effective execution of a large scale of sophisticated business processes. Nowadays, BPM systems have been deployed by many enterprises for a variety of applications such as manufacturing, service booking, supply chain management, etc. To some degree, the quality of business processes can determine the effectiveness of the enterprises. In reality, the processing of business process is always subject to the usage of available resources. It is thus essential to develop advanced resource management techniques for providing high-performance business process execution solutions and high quality workflow based services. So far, many efforts have been made to investigate this research problem. Conventional workflow scheduling approaches seek to optimise resource allocation at the run time. They are mainly used to deal with the applications in dynamic resource environment. However in many applications, business processes may have certain compulsory business requirements and goals to achieve in their execution. In such cases, business requirements cannot be well supported by the conventional workflow scheduling approaches sometimes. To handle this problem, resource planning for business processes must be conducted before execution, so that the business requirements and goals can be guaranteed to be eventually satisfied. In this thesis, we target to address the limitations of conventional workflow scheduling approaches by means of build time business process planning. Compared with previous works, our approach supports the compliance between business process execution and business level requirements. In order to perform robustly and effectively, build time resource planning for business processes must consider some particular features and constraints derived from the application scenarios. Firstly, in many cases the input structure of business process is sub-optimal, and this may cause resources unable to be optimally used; secondly, as the resource allocation is made at the build time, the scale of process instances for planning can be very large, and it will cause the problem to be computationally hard; thirdly, when we plan the future use of resources, the planning result must be consistent with the resource availability constraints; last but not the least, useful information is expected to be returned if the execution cannot comply with the given business requirements. To handle all the above issues, we propose some new concepts and develop a set of innovative algorithms for effective and efficient process planning. Specifically, we first investigate the issue of incorporating structural improvement into resource allocation. A novel approach is presented to evolve the process structure to cater for better resource utilisation. It enables the process structure and resource allocation to be tune with each other and hence better performance can be achieved than conventional approaches when the input process structure is un-optimised. Secondly, we deal with the applications in which the scale of process instances for planning is huge. In such cases, this problem is computationally hard because of the vast combination between tasks and resources. A novel batching technique is developed to batch the process instances, so that resource allocation can be carried out in a smaller search space. Also, we use innovative heuristics to understand the input factors and thus to facilitate efficient resource planning. If the business requirements (e.g. deadline) cannot be satisfied by the planning, we explore useful information for enterprise decision making and guidelines for negotiation with customers to rescue the business deals. Lastly, we plan the schedule of process instances in applications that resources have availability constraints. Three efficient novel approaches are introduced to schedule the instances in different criterions, and each of them may be superior in certain scenarios. Furthermore, we discuss how to extract some meaningful features from the test cases for classification, and the selectivity of suitable strategy for different applications. The major contributions of this research are that we have provided a set of new concepts, innovative methods and algorithms for resource planning for business processes at the build time. With these, optimised utilisation of available resources can be achieved, and the business goals and requirements can be guaranteed to be satisfied in the business process execution. As indicated by the experimental results, by deploying the set of proposed novel algorithms and techniques, the planning can be carried out in an effective and efficient way.