An Approach Based on Process Mining Techniques to Support Software Development

Abstract

In the software development lifecycle (SDLC) testing phase, extensive testing is performed on the code and programming to confirm that the software system operates according to the customer's requirements. The testing phase results are stored in event logs at system run-time. Process mining techniques aim to discover processes, verify conformance, and improve processes, by analyzing the instances contained in the event logs. Process mining allows analyzing a process automatically and determining where deviations are produced to take corrective actions. In this paper, an approach based on process mining techniques is proposed to support the testing phase in software development. The approach consists of a framework and a software tool that implements and automates its phases. Through applying the framework phases, a complete process model is discovered. Furthermore, implementing a clustering algorithm at the case level enables discovering sub-models of processes, representing the different behaviors identified in the event log. The framework also incorporates three well-known quality dimensions in process discovery (precision, recall, and generalization) to evaluate the quality of the discovered process models. Finally, a set of indicators are built for knowing the duplicated tasks, the average time consumed between transitions, activity sequences per case, and the time consumed, which allow identifying bottlenecks in the process software. The framework is evaluated using real-life event logs that confirm its effectiveness and efficiency. The proposed approach can complement the tools used in the testing phase, expecting that an improvement in the quality of the development process results in a better quality of the software product.