Data Mining for Process Modeling: A Clustered Process Discovery Approach

In this work, we explore an approach to process discovery that is based on combining several existing process discovery algorithms. We focus on algorithms that generate process models in the process tree notation, which are sound by design. The main components of our proposed process discovery approach are the Inductive Miner, the Evolutionary Tree Miner, the Local Process Model Miner and a new bottom-up recursive technique. We conjecture that the combination of these process discovery algorithms can mitigate some of the weaknesses of the individual algorithms. In cases where the Inductive Miner results in overgeneralizing process models, the Evolutionary Tree Miner can often mine much more precise models. At the other hand, while the Evolutionary Tree Miner is computationally expensive, running it only on parts of the log that the Inductive Miner is not able to represent with a precise model fragment can considerably limit the search space size of the Evolutionary Tree Miner. Local Process Models and bottom-up recursion aid the Evolutionary Tree Miner further by instantiating it with frequent process model fragments. We evaluate our approaches on a collection of real-life event logs and find that it does combine the advantages of the miners and in some cases surpasses other discovery techniques.

Healthcare processes are inherently complex as each patient is unique and medical staff deviate from protocols, often for valid reasons. Event logs collected by modern process-aware (healthcare) information systems provide a wealth of data and can be used to analyze the adherence to these protocols. Process mining is a young research area combining data science (machine learning, data mining, etc.) and business process management. Its main contributions have been techniques for process discovery (the automatic learning of process models from event data) and conformance checking (aligning observed and modeled behavior). However, existing techniques face challenging issues discovering high-quality process models in a healthcare setting. In this chapter we introduce the key concepts of process mining such as event logs, process models, and process discovery. We then show the application of two recent process mining techniques on a public event data set from the healthcare domain to demonstrate how some of the common pitfalls can be overcome. A first presented technique projects data statistics on a process model, allowing the analysis of correlations between patient characteristics and the executed activities (e.g., type of treatment). A second technique analyzes process performance without a specific need for a process model by considering contextual information. It correlates process characteristics with process performance.

Process mining techniques aim to analyze and improve conformance and performance of processes using event data. Process discovery is the most prominent process-mining task: A process model is derived based on an event log. The process model should be able to capture causalities, choices, concurrency, and loops. Process discovery is very challenging because of trade-offs between fitness, simplicity, precision, and generalization. Note that event logs typically only hold example behavior and cannot be assumed to be complete (to avoid overfitting). Dozens of process discovery techniques have been proposed. These use a wide range of approaches, e.g., language- or state-based regions, genetic mining, heuristics, expectation maximization, iterative log-splitting, etc. When models or logs become too large for analysis, the event log may be automatically decomposed or traces may be clustered before discovery. Clustering and decomposition are done automatically, i.e., no additional information is used. This paper proposes a different approach where a localized event log is assumed. Events are localized by assigning a non-empty set of regions to each event. It is assumed that regions can only interact through shared events. Consider for example the mining of software systems. The events recorded typically explicitly refer to parts of the system (components, services, etc.). Currently, such information is ignored during discovery. However, references to system parts may be used to localize events. Also in other application domains, it is possible to localize events, e.g., communication events in an organization may refer to multiple departments (that may be seen as regions). This paper proposes a generic process discovery approach based on localized event logs. The approach has been implemented in ProM and experimental results show that location information indeed helps to improve the quality of the discovered models.KeywordsVisible TraceConformance CheckVertical DecompositionInductive MinerUnique TraceThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Process discovery aims to derive a process model that accurately represents the observed behavior in an event log. As a state-of-the-art process discovery technique, Inductive Miner (IM) generates sound process models (i.e., free of deadlocks) while ensuring optimal replay fitness. However, IM may sometimes produce over-generalized process models with locally imprecise structures, often resulting in the creation of so-called flower structures. To address this limitation, this paper presents a novel technique that refines the process model generated by IM by optimizing its imprecise sub-processes. Specifically, the technique begins by identifying and extracting sub-logs corresponding to imprecise sub-processes in the initial IM-generated process model. Then, these imprecise sub-processes are iteratively optimized using a frequency-based filtering mechanism applied to the sub-logs. Once optimized, the imprecise sub-processes in the initial process model are replaced by the optimized ones, generating a set of candidates process models. Finally, the candidate with the best quality, in terms of fitness and precision, is selected as the final optimized process models. The proposed technique has been implemented as a plugin for the open-source process mining platform ProM. Through comparisons with state-of-the-art process discovery techniques using 10 publicly available real-life event logs, the experimental results demonstrate that the proposed method achieves an average absolute improvement of 0.173 in F-measure over its IMi variant, while also exhibiting competitive performance relative to other state-of-the-art approaches.

In this article we introduce <inline-formula><tex-math notation="LaTeX">${{\sf STARDUST}}$</tex-math></inline-formula> (event STream Analysis for pRocess Discovery Using Sampling sTragies), a process discovery approach that analyses a trace stream, in order to discover a process model that may change over time. The basic idea is to adopt a sampling technique to select the most representative trace variants to be considered for the process discovery, then to alert a concept drift as the trace variants to be sampled change over time and, finally, to trigger the discovery of a new process model as a drift is alerted. We formulate the proposed approach under the assumption that the trace distribution commonly follows the Pareto's principle (i.e., a few trace variants covers the majority of cases) which is commonly satisfied in several business processes. Experimental results on various benchmark event logs handled as streams show the effectiveness of the proposed approach also compared to a state-of-the- art concept drift detection approach.

Process mining techniques have matured over the last decade and more and more organization started to use this new technology. The two most important types of process mining are process discovery (i.e., learning a process model from example behavior recorded in an event log) and conformance checking (i.e., comparing modeled behavior with observed behavior). Process mining is motivated by the availability of event data. However, as event logs become larger (say terabytes), performance becomes a concern. The only way to handle larger applications while ensuring acceptable response times, is to distribute analysis over a network of computers (e.g., multicore systems, grids, and clouds). This paper provides an overview of the different ways in which process mining problems can be distributed. We identify three types of distribution: replication, a horizontal partitioning of the event log, and a vertical partitioning of the event log. These types are discussed in the context of both procedural (e.g., Petri nets) and declarative process models. Most challenging is the horizontal partitioning of event logs in the context of procedural models. Therefore, a new approach to decompose Petri nets and associated event logs is presented. This approach illustrates that process mining problems can be distributed in various ways.

Background: Standard operating procedure (SOP) is a series of business activities to achieve organisational goals, with each activity carried to be recorded and stored in the information system together with its location (e.g., SCM, ERP, LMS, CRM). The activity is known as event data and is stored in a database known as an event log.Objective: Based on the event log, we can calculate the fitness to determine whether the business process SOP is following the actual business process.Methods: This study obtains the event log from a terminal operating system (TOS), which records the dwelling time at the container port. The conformance checking using token-based replay method calculates fitness by comparing the event log with the process model.Results: The findings using the Alpha algorithm resulted in the most traversed traces (a, b, n, o, p). The fitness calculation returns 1.0 were produced, missing, and remaining tokens are replied to each of the other traces.Conclusion: Thus, if the process mining produces a fitness of more than 0.80, this shows that the process model is following the actual business process. Keywords: Conformance Checking, Dwelling time, Event log, Fitness, Process Discovery, Process Mining

Process mining facilitates the discovery, conformance, and enhancement of business processes using event logs. However, incomplete event logs and the complexities of concurrent activities present significant challenges in achieving accurate process models that fulfill the completeness condition required in process mining. This paper introduces a Timed Genetic-Inductive Process Mining (TGIPM) algorithm, a novel approach that integrates the strengths of Timed Genetic Process Mining (TGPM) and Inductive Mining (IM). TGPM extends traditional Genetic Process Mining (GPM) by incorporating time-based analysis, while the IM is widely recognized for producing sound and precise process models. For the first time, these two algorithms are combined into a unified framework to address both missing activity recovery and structural correctness in process discovery. This study evaluates two scenarios: a sequential approach, in which TGPM and IM are executed independently and sequentially, and the TGIPM approach, where both algorithms are integrated into a unified framework. Experimental results using real-world event logs from a health service in Indonesia demonstrate that TGIPM achieves higher fitness, precision, and generalization compared to the sequential approach, while slightly compromising simplicity. Moreover, the TGIPM algorithm exhibits lower computational cost and more effectively captures parallelism, making it particularly suitable for large and incomplete datasets. This research underscores the potential of TGIPM to enhance process mining outcomes, offering a robust framework for accurate and efficient process discovery while driving process innovation across industries.

Process mining enables organizations to uncover their actual processes, provide insights, diagnose problems, and automatically trigger corrective actions. Process mining is an emerging scientific discipline positioned at the intersection between process science and data science. The combination of process modeling and analysis with the event data present in today’s information systems provides new means to tackle compliance and performance problems. This chapter provides an overview of the field of process mining introducing the different types of process mining (e.g., process discovery and conformance checking) and the basic ingredients, i.e., process models and event data. To prepare for later chapters, event logs are introduced in detail (including pointers to standards for event data such as XES and OCEL). Moreover, a brief overview of process mining applications and software is given.

A marked Petri net is lucent if there are no two different reachable markings enabling the same set of transitions, i.e., states are fully characterized by the transitions they enable. This paper explores the class of marked Petri nets that are lucent and proves that perpetual marked free-choice nets are lucent. Perpetual free-choice nets are free-choice Petri nets that are live and bounded and have a home cluster, i.e., there is a cluster such that from any reachable state there is a reachable state marking the places of this cluster. A home cluster in a perpetual net serves as a “regeneration point” of the process, e.g., to start a new process instance (case, job, cycle, etc.). Many “well-behaved” process models fall into this class. For example, the class of short-circuited sound workflow nets is perpetual. Also, the class of processes satisfying the conditions of the \(\alpha \) algorithm for process discovery falls into this category. This paper shows that the states in a perpetual marked free-choice net are fully characterized by the transitions they enable, i.e., these process models are lucent. Having a one-to-one correspondence between the actions that can happen and the state of the process, is valuable in a variety of application domains. The full characterization of markings in terms of enabled transitions makes perpetual free-choice nets interesting for workflow analysis and process mining. In fact, we anticipate new verification, process discovery, and conformance checking techniques for the subclasses identified.

Event data are collected everywhere: in logistics, manufacturing, finance, healthcare, e-learning, e-government, and many other domains. The events found in these domains typically refer to activities executed by resources at particular times and for particular cases. Process mining provides the means to discover the real processes, to detect deviations from normative processes, and to analyze bottlenecks and waste from such events. However, process mining tends to be backward-looking. Fortunately, simulation can be used to explore different design alternatives and to anticipate future performance problems. This keynote paper discusses the link between both types of analysis and elaborates on the challenges process discovery techniques are facing. Quality notions such as recall, precision, and generalization are discussed. Rather than introducing a specific process discovery or conformance checking algorithm, the paper provides a comprehensive set of conformance propositions. These conformance propositions serve two purposes: (1) introducing the essence of process mining by discussing the relation between event logs and process models, and (2) discussing possible requirements for the quantification of quality notions related to recall, precision, and generalization.

Process mining is an emerging discipline incorporating methods and tools for the analysis of system/process executions captured in the form of event logs [1]. Traditionally process mining can be divided into three research areas: discovery (construction of process models from event logs), conformance checking (finding log and model deviations), and enhancement of existing process models with additional event log data.BPMN (Business Process Model and Notation) 2.0 [2] is a widely used process modeling notation, supported by various process modeling and analysis tools, and is a de-facto process modeling standard. Using BPMN within process mining opens perspectives for applicability of the existing process mining techniques: for instance discovered process models can be analyzed or enacted using existing BPMN-based software, and vice versa, manually created models can be imported to a process mining tool, verified against event logs, and enhanced with additional data.In this work we bridge the gap between conventional process modeling formalisms used in the context of process mining (e.g., Petri nets, causal nets, process trees) and BPMN. For that purpose we developed a suite of conversion algorithms and provide formal guarantees relating the behavior of Petri nets (including non-free-choice nets) to the corresponding BPMN models (and vice versa). The derived relations are used to enhance the BPMN models with information learned from the event logs.The developed conversion techniques are described in detail in [3] and have been implemented [4] as a part of ProM (Process Mining Framework) [5] -- an open source tool for process mining and verified on real event log data. Moreover, cases for which conversion algorithms give more compact process models in comparison with the initial models are identified. Although the developed algorithms deal with basic control flow constructs only, they can be applied in the discovery of advanced BPMN modeling elements [2], including subprocesses [6-7], cancellations [8], conditional branching and data objects, swimlanes, message flows, and others.

Many organisations now seek to analyse and improve their processes using event logs from various IT systems supporting their operations. Process mining aims to obtain insights from such process data, using process discovery, conformance checking and performance measures. While many commercial process mining tools feature user-friendly directly follows-based process maps, they typically do not offer a way to assess the quality of the model, leaving users with potentially unreliable insights, which could lead to the wrong conclusion being drawn from these insights. In contrast, academic tools typically provide verifiable results, but are often difficult to use and understand for stakeholders, sometimes overgeneralising behaviour to fit more extensive process model formalisms. In this paper, we bridge this well-known gap between commercial and academic tools by combining sound process discovery, conformance checking and performance capabilities with user-friendly directly follows-based process models. We implemented these techniques in a new process mining tool and applied them to analyse several business processes in a Queensland Government department. We discovered sound directly follows-based process models from their logs, compared them with prescribed models and analysed the performance of these processes. In particular, our conformance checking techniques allowed to pinpoint deviations between prescribed processes and actual recorded behaviour. The outcomes of this case study are now being used to document, review, improve and automate processes.

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.

Process discovery aims at building process models using information retrieved from logs. Process characteristics play a significant role in the selection of a suitable process modeling language for describing process discovery results. Business processes characterized by high variability, in which participants have a lot of autonomy and flexibility in executing the process, are difficult to be described with procedural process modeling languages,since they explicitly represent in a model every possible path. Declarative languages, like Declare, alleviate this issue by defining a set of constraints between activities that must not be violated during the process execution instead of describing what to do step by step. Recently, several process discovery techniques have been proposed for extracting a set of Declare constraints from a log. However, no one of these techniques allows the user to exploit the time perspective often available in a log to discover “time-aware” Declare constraints. Timed Declare has already previously been introduced to monitor metric temporal constraints at runtime. In this paper, we use this semantics for discovering a set of Timed Declare constraints from an event log. We have implemented the proposed approach as a plug-in of the process mining tool ProM. We have validated the approach by using our plug-in to mine two real-life event logs.