Sebastiaan J. van Zelst

Avoiding Over-Fitting in ILP-Based Process Discovery

The aim of process discovery is to discover a process model based on business process execution data, recorded in an event log. One of several existing process discovery techniques is the ILP-based process discovery algorithm. The algorithm is able to unravel complex process structures and provides formal guarantees w.r.t. the model discovered, e.g., the algorithm guarantees that a discovered model describes all behavior present in the event log. Unfortunately the algorithm is unable to cope with exceptional behavior present in event logs. As a result, the application of ILP-based process discovery techniques in everyday process discovery practice is limited. This paper addresses this problem by proposing a filtering technique tailored towards ILP-based process discovery. The technique helps to produce process models that are less over-fitting w.r.t. the event log, more understandable, and more adequate in capturing the dominant behavior present in the event log. The technique is implemented in the ProM framework.

Improving Process Discovery Results by Filtering Outliers Using Conditional Behavioural Probabilities

Process discovery, one of the key challenges in process mining, aims at discovering process models from process execution data stored in event logs. Most discovery algorithms assume that all data in an event log conform to correct execution of the process, and hence, incorporate all behaviour in their resulting process model. However, in real event logs, noise and irrelevant infrequent behaviour are often present. Incorporating such behaviour results in complex, incomprehensible process models concealing the correct and/or relevant behaviour of the underlying process. In this paper, we propose a novel general purpose filtering method that exploits observed conditional probabilities between sequences of activities. The method has been implemented in both the ProM toolkit and the RapidProM framework. We evaluate our approach using real and synthetic event data. The results show that the proposed method accurately removes irrelevant behaviour and, indeed, improves process discovery results.

Event stream-based process discovery using abstract representations

The aim of process discovery, originating from the area of process mining, is to discover a process model based on business process execution data. A majority of process discovery techniques relies on an event log as an input. An event log is a static source of historical data capturing the execution of a business process. In this paper, we focus on process discovery relying on online streams of business process execution events. Learning process models from event streams poses both challenges and opportunities, i.e. we need to handle unlimited amounts of data using finite memory and, preferably, constant time. We propose a generic architecture that allows for adopting several classes of existing process discovery techniques in context of event streams. Moreover, we provide several instantiations of the architecture, accompanied by implementations in the process mining toolkit ProM (http://promtools.org). Using these instantiations, we evaluate several dimensions of stream-based process discovery. The evaluation shows that the proposed architecture allows us to lift process discovery to the streaming domain.

Online conformance checking: relating event streams to process models using prefix-alignments

Companies often specify the intended behaviour of their business processes in a process model. Conformance checking techniques allow us to assess to what degree such process models and corresponding process execution data correspond to one another. In recent years, alignments have proven extremely useful for calculating conformance checking statistics. Existing techniques to compute alignments have been developed to be used in an offline, a posteriori setting. However, we are often interested in observing deviations at the moment they occur, rather than days, weeks or even months later. Hence, we need techniques that enable us to perform conformance checking in an online setting. In this paper, we present a novel approach to incrementally compute prefix-alignments, paving the way for real-time online conformance checking. Our experiments show that the reuse of previously computed prefix-alignments enhances memory efficiency, whilst preserving prefix-alignment optimality. Moreover, we show that, in case of computing approximate prefix-alignments, there is a clear trade-off between memory efficiency and approximation error.

Filtering Spurious Events from Event Streams of Business Processes

Process mining aims at gaining insights into business processes by analysing event data recorded during process execution. The majority of existing process mining techniques works offline, i.e. using static, historical data stored in event logs. Recently, the notion of online process mining has emerged, whereby techniques are applied on live event streams, as process executions unfold. Analysing event streams allows us to gain instant insights into business processes. However, current techniques assume the input stream to be completely free of noise and other anomalous behaviours. Hence, applying these techniques to real data leads to results of inferior quality. In this paper, we propose an event processor that enables us to filter out spurious events from a live event stream. Our experiments show that we are able to effectively filter out spurious events from the input stream and, as such, enhance online process mining results.

Repairing Alignments: Striking the Right Nerve

Process Mining is concerned with the analysis, understanding and improvement of business processes. One of the most important branches of process mining is conformance checking, i.e. assessing to what extent a business process model conforms to observed business process execution data. Alignments are the de facto standard instrument to compute conformance statistics. Alignments map elements of an event log onto activities present in a business process model. However, computing them is a combinatorial problem and hence, extremely costly. In this paper we show how to compute an alignment for a given process model, using an existing alignment and an existing process model as a basis. We show that we are able to effectively repair the existing alignment by updating those parts that no longer fit the given process model. Thus, computation time decreases significantly. Moreover, we show that the potential loss of optimality is limited and stays within acceptable bounds.

Repairing outlier behaviour in event logs

One of the main challenges in applying process mining on real event data, is the presence of noise and rare behaviour. Applying process mining algorithms directly on raw event data typically results in complex, incomprehensible, and, in some cases, even inaccurate analyses. As a result, correct and/or important behaviour may be concealed. In this paper, we propose an event data repair method, that tries to detect and repair outlier behaviour within the given event data. We propose a probabilistic method that is based on the occurrence frequency of activities in specific contexts. Our approach allows for removal of infrequent behaviour, which enables us to obtain a more global view of the process. The proposed method has been implemented in both the ProM- and the RapidProM framework. Using these implementations, we conduct a collection of experiments that show that we are able to detect and modify most types of outlier behaviour in the event data. Our evaluation clearly demonstrates that we are able to help to improve process mining discovery results by repairing event logs upfront.

Online Conformance Checking Using Behavioural Patterns

New and compelling regulations (e.g., the GDPR in Europe) impose tremendous pressure on organizations, in order to adhere to standard procedures, processes, and practices. The field of conformance checking aims to quantify the extent to which the execution of a process, captured within recorded corresponding event data, conforms to a given reference process model. Existing techniques assume a post-mortem scenario, i.e. they detect deviations based on complete executions of the process. This limits their applicability in an online setting. In such context, we aim to detect deviations online (i.e., in-vivo), in order to provide recovery possibilities before the execution of a process instance is completed. Also, current techniques assume cases to start from the initial stage of the process, whereas this assumption is not feasible in online settings. In this paper, we present a generic framework for online conformance checking, in which the underlying process is represented in terms of behavioural patterns and no assumption on the starting point of cases is needed. We instantiate the framework on the basis of Petri nets, with an accompanying new unfolding technique. The approach is implemented in the process mining tool ProM, and evaluated by means of several experiments including a stress-test and a comparison with a similar technique.

Maximizing Synchronization for Aligning Observed and Modelled Behaviour

Conformance checking is a branch of process mining that aims to assess to what degree event data originating from the execution of a (business) process and a corresponding reference model conform to each other. Alignments have been recently introduced as a solution for conformance checking and have since rapidly developed into becoming the de facto standard.

Applying Sequence Mining for Outlier Detection in Process Mining

One of the challenges in applying process mining algorithms on real event data, is the presence of outlier behavior. Such behaviour often leads to complex, incomprehensible, and, sometimes, even inaccurate process mining results. As a result, correct and/or important behaviour of the process may be concealed. In this paper, we exploit sequence mining techniques for the purpose of outlier detection in the process mining domain. Using the proposed approach, it is even possible to detect outliers in case of heavy parallelism and/or long-term dependencies between business process activities. Our method has been implemented in both the ProM- and the RapidProM framework. Using these implementations, we conducted a collection of experiments that show that we are able to detect and remove outlier behaviour in event data. Our evaluation clearly demonstrates that the proposed method accurately removes outlier behaviour and, indeed, improves process discovery results.