Conformance checking, a core activity in process mining, compares real-world process executions recorded in event logs to intended process behavior captured in process models. While numerous conformance checking techniques exist, their practical adoption remains limited. One contributing factor to this is that visualizations might not be suited to address the underlying user purposes. In that light, recent work has introduced a taxonomy of conformance checking tasks that captures purposes when applying conformance checking. However, current tools neither systematically evaluate their visualizations against this taxonomy nor provide integrated support for end-to-end conformance analysis. To address this gap, this paper presents CC Viz, a tool that ofers five targeted visualizations, each tailored to a specific task from the taxonomy. The tasks-ranging from exploring guideline violations to presenting the impact of conformance on process outcomes-are organized as a sequential analysis pipeline. CC Viz enables users to switch between visualizations and interactively drill down into specific conformance violations, facilitating a holistic and task-aligned conformance analysis experience.
CEUR
ceur-ws.org Source code repository Screencast video
Value CC Viz 1.0 Apache 2.0 https://github.com/michaelgrohs/ccviz/blob/master/README.md https://github.com/michaelgrohs/ccviz https://github.com/michaelgrohs/ccviz/blob/master/demonstration.mov developed. They are able to detect conformance violations whenever a trace diverges from the process model [2]. All techniques require a process model and an event log as input [1]. Based on that, users can automatically derive non-conform behavior and also directly assess which parts of traces were not desired [3].
Regardless of these capabilities, conformance checking is rarely applied in practice [4, p. 39]. This can be attributed to multiple reasons. For example, many conformance checking techniques require significant computational resources, leading to long waiting times during their usage [ 5]. Another reason is the reliance on process models, which may not be available in organizations since it is time-consuming and error-prone to create and maintain them [2].
In this demo, we address another reason for the lack of conformance checking usage in practice: the suitability of visualizations to address users’ needs when applying conformance checking. These needs are the purpose that a user has when using conformance checking, or, in other words, the task the user wants to fulfill [ 5]. Recently, a taxonomy of conformance checking tasks has been proposed, summarizing manifold purposes of conformance checking [5]. The taxonomy consists of six dimensions: • task goal: why the task is done (e.g., explore, confirm) • task means: how the task is carried out (e.g., discover, compare) • data characteristics: what should be revealed (e.g., conformance, guideline violations1) • constraint type: the perspective referred to (e.g., control-flow, data) • data target: the data on which the task is carried out (e.g., log, trace) • data cardinality: the cardinality of the data target (e.g., all, many) Through this, users are able to communicate what they want to achieve with conformance checking. However, existing visualizations in commercial or scientific tools are not evaluated w.r.t. whether they address these tasks. Additionally, existing tools predominantly contain standalone visualizations, meaning that there is no support for a wholistic conformance analysis.
To address this gap and closely align the visualization of conformance checking results with
the underlying tasks, we present the CC Viz tool. Given a process model and an event log as
input, the tool visualizes conformance checking results that address five common tasks of the
dimensions ‘task goal: task means, data characteristics’:
(
In the remainder of this paper, we introduce the CC Viz tool with all its functionalities in section 2, after which we illustrate its maturity in section 3 and conclude in section 4. 1Note that guideline violations most commonly refer to deviations from a process model [5].
At https://github.com/michaelgrohs/ccviz, the CC Viz tool can be accessed. In this repository, the user can find the source code, instructions on how to run the tool, and further documentation. A demo video is available at https://github.com/michaelgrohs/ccviz/blob/master/demonstration. mov. Alternatively, it is also possible to access the tool in a hosted setting without requiring any local setup via https://ccviz-frontend.vercel.app. This relies on a hosted backend service Render (https://render.com) and a frontend service Vercel (https://vercel.com). Note that performance is significantly slower than the local setup and the backend service is sleeping per default. Thus, the functionalities are not available directly and has to be woken up. Also, the service restricts RAM to 512 MB, meaning that it cannot handle larger files (the default dataset works).
As illustrated in Fig. 1, a process model and an event log are required as input for CC Viz. Then,
all relevant computations are performed simultaneously for all tasks once, meaning that no
waiting times are required after that. This computation uses current state-of-the-art
alignmentbased conformance checking as basis which provides detailed feedback on skipped and inserted
activities per trace as well as fitness levels. Following that, the user enters the task-specific
visualizations, starting with task (
(
(
Conformance distribution
(
In CC Viz, a Python-based back-end and a React-based front-end communicate through request and response mechanisms. The back-end utilizes Flask to ensure communication even if the front-end is closed. Further, PM4Py’s trace alignments [7] handle the conformance check.
We used the tool for the processes of the BPI Challenges 2012 (sub-process with A_ and O_ activities only; 12A & 12O) and 2020 (Domestic Declarations (Dom.), International Declarations (Int.), Request for Payment (RfP), and Prepaid Travel Costs (Prep.)) obtained from [6] as well as the sepsis process [8] and the road trafic fines process obtained from [ 5]. These processes represent a variety of examples with respect to size, complexity, and count of recorded attributes.
Tab. 1 shows descriptive statistics for all used event log-process model pairs and required
computation times. We see that CC Viz is heavily impacted by the computation of alignments,
which is time-intensive for processes with many, long variants and complex models like Sepsis
with 354 seconds (about 5.8 minutes). Road trafic also takes relatively long due to its size with
122 seconds (about 2.0 minutes). All other processes take less than 30 seconds. This indicates
that the attribute processing for task (
Computation
Time
We presented CC Viz, a tool for the visualization of conformance checking results. It aims to provide visualizations that are tailored exactly to the purpose that users have when applying conformance checking. To achieve that, the visualizations are based on tasks identified in a corresponding taxonomy [5] and accessible in a common sequential analysis pipeline. In the future, we aim to empirically assess whether users can use CC Viz to solve the underlying tasks.
The authors used ChatGPT, DeepL for: Grammar & spelling check, Paraphrase & reword. The authors reviewed and edited the content as needed and take full responsibility for all content.