Discovering Context-Aware Models for Predicting Business Process Performances

Abstract

Abstract. Discovering predictive models for run-time support is an emerging topic in Process Mining research, which can effectively help optimize business process enactments. However, making accurate estimates is not easy especially when considering fine-grain performance measures (e.g., processing times) on a complex and flexible business process, where performance patterns change over time, depending on both case properties and context factors (e.g., seasonality, workload). We try to face such a situation by using an ad-hoc predictive clustering approach, where different context-related execution scenarios are discovered and modeled accurately via distinct state-aware performance predictors. A readable predictive model is obtained eventually, which can make performance forecasts for any new running process case, by using the predictor of the cluster it is estimated to belong to. The approach was implemented in a system prototype, and validated on a real-life context. Test results confirmed the scalability of the approach, and its efficacy in predicting processing times and associated SLA violations.

A recommendation system for predicting risks across multiple business process instances

“…Transition systems are employed to provide predictions using similarly completed executions as a reference. Folino et al [26] use a combination of clustering techniques and transition systems. Using clustering they identify process variants in a log and for each cluster they generate a transition system.…”

Queue mining for delay prediction in multi-class service processes

“…To encompass context factors into the state abstraction, Folino et al [16] extended the work on performance prediction based on transition systems and defined a state abstraction that comprised of two types of features: (1) 'internal' case properties and (2) 'external' factors that characterize system state, e.g. workload, resource availability.…”

“…delays, remaining times) and the limitation to decision trees as learning techniques (we discuss these limitations in more detail when reviewing related work). Therefore, in this paper, we undertake a more flexible approach that enables the use of various types of learning techniques and combines transition systems that comprise of cases and case-specific attributes (similarly to the internal features of Folino et al [16]) with continuous vectors of system-state factors. This results in a decoupling of the state (which remains simple and case-related) from the complex (and possibly continuous) feature vectors when applying the learning technique.…”

“…In this work, we also consider the more complex setup of multi-class settings. To make the regression-based approach work for the multi-class setting, we also propose an extension to a prediction method based on transition systems developed by van der Aalst et al [4] and enhanced with context factors by Folino et al [16]. In particular, we show how a combination of characteristics of cases and the overall system state can be incorporated following a machine learning approach.…”

A General Framework for Correlating Business Process Characteristics

“…Many authors have proposed techniques to relate specific characteristics in an ad-hoc manner. For example, several approaches have been proposed to predict the remaining processing time of a case depending on characteristics of the partial trace executed [1][2][3]. Other approaches are only targeted to correlating certain predefined characteristics to the process outcome [4][5][6] or the violations of business rules [7].…”