A process mining oriented approach to improve process models analysis in developing countries

As a result of the independence and the colonization endured by developing countries, most of their administrative procedures have been inherited from the colonial era. These procedures adapted to the colonizing countries, are complex for the African context. People in charge of their processing are not able to master it, take too much time to put them in practice and consider it as heavy for the administration. Sometimes, some misbehaviors are introduced in the execution of processes to accelerate it; this causes many troubles and incomprehension in their analysis, leading to bad decisions-making within the organization; It is therefore imperative to detect and solve these misbehaviors. The resulting challenge is to improve the comprehension of descriptive process models executed within information systems of developing countries by reducing their structural complexity and detecting anomalies in their execution. A substantial amount of literature has been devoted to this topic, and classified it among the set of problems solved by process mining techniques. Process mining is a research discipline interested in the proposal of methods for extracting descriptive business process models incomprehensible by users. Several techniques have been proposed for this purpose among which the α-algorithm, central for this work. The aim of this paper is to propose a refinement of this algorithm in order to generate a less structurally complex and more understandable descriptive process from execution logs within an information system. This refinement is essentially based on splitting the process into sub-processes according to the set of roles involved in the business process execution. The proposed model is experimented on a gynecological oncology workflow log and, it is shown how the whole process is split into sub-processes leading to a saving of execution time equivalent to half the execution time of the basic model.