Robotic Process Mining

Robotic Process Automation (RPA) has lowered the threshold for process automation. Repetitive tasks done by people are handed over to software robots. For RPA, there is no need to change or replace the pre-existing information systems. Instead, software robots replace users by interacting directly with the user interfaces normally operated by humans. Actually, RPA can be seen as "the poor man's workflow management solution" because it is cheaper than traditional automation. Therefore, it can be used to automate routine work that would normally not be cost-effective. Process mining plays a key role in deciding what to automate and how. Therefore, RPA is closely related to process mining. Before introducing RPA, one needs to analyze the processes to be automated. Process mining can help to identify promising candidates. Moreover, after RPA has been implemented, process mining can be used to monitor processes and systems even if these use a mixture of RPA, workers, and traditional automation.

Process mining techniques analyze processes based on events stored in event logs. Yet, low-level events recorded by information systems may not directly match high-level activities that make sense to process stakeholders. This results in discovered process models that cannot be easily understood. To prevent such situations from happening, low-level events need to be translated into highlevel activities that are recognizable by stakeholders. This paper proposes the Guided Process Discovery method (GPD). Low-level events are grouped based on behavioral activity patterns, which capture domain knowledge on the relation between high-level activities and low-level events. Events in the resulting abstracted event log correspond to instantiations of high-level activities. We validate process models discovered on the abstracted event log by checking conformance between the low-level event log and an expanded model in which the high-level activities are replaced by activity patterns. The method was tested using two real-life event logs. We show that the process models discovered with the GPD method are more comprehensible and can be used to answer process questions, whereas process models discovered using standard process discovery techniques do not provide the insights needed.

Lecture Notes in Computer Science, 2009

Process mining refers to the extraction of process models from event logs. Real-life processes tend to be less structured and more flexible. Traditional process mining algorithms have problems dealing with such unstructured processes and generate spaghetti-like process models that are hard to comprehend. One reason for such a result can be attributed to constructing process models from raw traces without due pre-processing. In an event log, there can be instances where the system is subjected to similar execution patterns/behavior. Discovery of common patterns of invocation of activities in traces (beyond the immediate succession relation) can help in improving the discovery of process models and can assist in defining the conceptual relationship between the tasks/activities. In this paper, we characterize and explore the manifestation of commonly used process model constructs in the event log and adopt pattern definitions that capture these manifestations, and propose a means to form abstractions over these patterns. We also propose an iterative method of transformation of traces which can be applied as a pre-processing step for most of today's process mining techniques. The proposed approaches are shown to identify promising patterns and conceptually-valid abstractions on a real-life log. The patterns discussed in this paper have multiple applications such as trace clustering, fault diagnosis/anomaly detection besides being an enabler for hierarchical process discovery.

IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2013

There are many process mining algorithms and representations, making it difficult to choose which algorithm to use or compare results. Process mining is essentially a machine learning task, but little work has been done on systematically analyzing algorithms to understand their fundamental properties, such as how much data are needed for confidence in mining. We propose a framework for analyzing process mining algorithms. Processes are viewed as distributions over traces of activities and mining algorithms as learning these distributions. We use probabilistic automata as a unifying representation to which other representation languages can be converted. We present an analysis of the Alpha algorithm under this framework and experimental results, which show that from the substructures in a model and behavior of the algorithm, the amount of data needed for mining can be predicted. This allows efficient use of data and quantification of the confidence which can be placed in the results.

IPA - AAAI Workshop, 2020

Robotic Process Automation (RPA) is a technology for automating repetitive routines consisting of sequences of user interactions with one or more applications. In order to fully exploit the opportunities opened by RPA, companies need to discover which specific routines may be automated, and how. In this setting, this paper addresses the problem of analyzing User Interaction (UI) logs in order to discover routines where a user transfers data from one spreadsheet or (Web) form to another. The paper maps this problem to that of discovering data transformations by example-a problem for which several techniques are available. The paper shows that a naive application of a state-of-the-art technique for data transformation discovery is computationally inefficient. Accordingly, the paper proposes two optimizations that take advantage of the information in the UI log and the fact that data transfers across applications typically involve copying alphabetic and numeric tokens separately. The proposed approach and its optimizations are evaluated using UI logs that replicate a real-life repetitive data transfer routine.

Anais do IV Simpósio Brasileiro de Sistemas de Informação (SBSI 2008), 2008

The paper describes FlowSpy, an environment that employs a sequence mining technique to discover and analyze actual process execution paths from business processes, for both process comparison and process discovery. FlowSpy focuses on exploratory analysis of the different execution flows, enabling a detailed analysis of business behavior, quantification of different execution flows, and abstraction mechanisms (log pre-processing and visualization abstraction) that deal with process complexity and different process views. Log pre-processing aims at improving the data mining phase, with a more restricted aggregate tree. Visualization abstraction facilitates pattern interpretation by producing trees that represent the obtained patterns.

Process mining is able to reveal how people and organizations really function. Often reality is very different and less structured than expected. Process discovery exposes the variability of real-life processes. Conformance checking is able to pinpoint and diagnose compliance problems. Task mining exploits user-interaction data to enrich traditional event data. All these different forms of process mining can and should support Robotic Process Automation (RPA) initiatives. Process mining can be used to decide what to automate and to monitor the cooperation between software robots, people, and traditional information systems. In the process of deciding what to automate, the Pareto principle plays an important role. Often 80% of the behavior in the event data is described by 20% of the trace variants or activities. An organization can use such insights to "pick its automation battles", e.g., analyzing the economic and practical feasibility of RPA opportunities before implementation. This paper discusses how to leverage the Pareto principle in RPA and other process automation initiatives.

Data & Knowledge Engineering, 2008

Process mining techniques have been receiving great attention in the literature for their ability to automatically support process (re)design. Typically, these techniques discover a concrete workflow schema modelling all possible execution patterns registered in a given log, which can be exploited subsequently to support further-coming enactments. In this paper, an approach to process mining is introduced that extends classical discovery mechanisms by means of an abstraction method aimed at producing a taxonomy of workflow models. The taxonomy is built to capture the process behavior at different levels of detail. Indeed, the most-detailed mined models, i.e., the leafs of the taxonomy, are meant to support the design of concrete workflows, as it happens with existing techniques in the literature. The other models, i.e., non-leaf nodes of the taxonomy, represent instead abstract views over the process behavior that can be used to support advanced monitoring and analysis tasks. All the techniques discussed in the paper have been implemented, tested, and made available as a plugin for a popular process mining framework (ProM). A series of tests, performed on different synthesized and real datasets, evidenced the capability of the approach to characterize the behavior encoded in input logs in a precise and complete way, achieving compelling conformance results even in the presence of complex behavior and noisy data. Moreover, encouraging results have been obtained in a real-life application scenario, where it is shown how the taxonomical view of the process can effectively support an explorative ex-post analysis, hinged on the different kinds of process execution discovered from the logs.

Conference on Advanced Information Systems Engineering, 2018

Robotic Process Automation (RPA) is a novel approach for immediate cost reduction and gaining operational efficiency. RPA tools can automate repeatable tasks, thus reducing the error rates and increasing overall process performance. Even more, RPA improves tne quality of the data (data completeness, data consistency/correctness, etc.). Although, being widely used in many organisations, RPA suffers from high time consumption allocated to the training of software robots (bots for short). Moreover, the models used for training are often inaccurate, which leads to increase of time spent on testing the bots. One of the possible solutions is to apply process mining in order to extract the information about the processes from UI logs such as clickstreams and keylogs, which can then be used to train the bots. However, traditional process discovery techniques are not suitable for the purpose of RPA, as they discover only control-flow perspective of the process and can not deal well with the UI logs, producing huge and complex models. The proposed research project aims at shifting process mining techniques from working on event logs to working on UI logs as well as developing multi-perspective automated discovery technique, which can than be applied to train the RPA bots.

Event data are collected in logistics, manufacturing, finance, healthcare, customer relationship management, 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 a particular case (i.e., process instances). Process mining techniques are able to exploit such data. In this article, we focus on process discovery. However, process mining also includes conformance checking, performance analysis, decision mining, organizational mining, predictions, recommendations, etc. These techniques help to diagnose problems and improve processes. All process mining techniques involve both event data and process models. Therefore, a typical first step is to automatically learn a control-flow model from the event data. This is very challenging, but in recent years many powerful discovery techniques have been developed. It is not easy to compare these techniques since they use different representations and make different assumptions. Users often need to resort to trying different algorithms in an ad-hoc manner. Developers of new techniques are often trying to solve specific instances of a more general problem. Therefore, we aim to unify existing approaches by focusing on log and model abstractions. These abstractions link observed and modeled behavior: Concrete behaviors recorded in event logs are related to possible behaviors represented by process models. Hence, such behavioral abstractions provide an "interface" between both. We discuss four discovery approaches involving three abstractions and different types of process models (Petri nets, block-structured models, and declarative models). The goal is to provide a comprehensive understanding of process discovery and show how to develop new techniques. Examples illustrate the different approaches and pointers to software are given. The discussion on abstractions and process representations is also used to reflect on the gap between process mining literature and commercial process mining tools. This facilitates users to select an appropriate process discovery technique. Moreover, structuring the role of internal abstractions and representations helps to broaden the view and facilitates the creation of new discovery approaches.