A Web Application Supporting the Full Pipeline of Business Process Deviance Analysis

Abstract

In business process mining, the deviant cases refer to the unusual cases in the process execution flow. Depending on their performance and outcomes, processes can deviate in negative ways (for example a delivery process that takes too much time) or positive ways (for example, a healthcare process in which a patient recovered very quickly). Business process deviance mining is the task of exploring the reasons behind exceptional cases in business process logs. In this thesis, we introduce a web application built on top of existing work concerning the problem of explaining deviant cases using sequential or declarative process patterns characterizing the cases, or a combination of them. While the existing work provided most of the backend of the application, we developed a web application on top of it to guide the process analyst in the deviance mining task throughout the entire analysis pipeline from log splitting, to case labeling, to the application of classifiers to extract deviance explanations in terms of process patterns. The development and design of our application bases on a set of requirements acquired from BPM experts. In this thesis, we will first present the requirements, then we will walk through how each requirement is fulfilled by our implementation by creating test cases for each specific requirement.