Event data, often stored in the form of event logs, serve as the starting point for process mining and other evidence-based process improvements. However, event data in logs are often tainted by noise, errors, and missing data. Recently, a novel body of research has emerged, with the aim to address and analyze a class of anomalies known as uncertainty-imprecisions quantified with meta-information in the event log. This paper illustrates an extension of the XES data standard capable of representing uncertain event data. Such an extension enables input, output, and manipulation of uncertain data, as well as analysis through the process discovery and conformance checking approaches available in literature.
Through the last decades, the increase in the availability of data generated by the execution of processes has enabled the development of the set of disciplines known as process sciences. These fields of science aim to analyze data accounting for the process perspective—the flow of events belonging to a process case.
Uncertain event data is a newly-emerging class of anomalous event data. Uncertain data
consists of events that have been logged with a quantified measure of uncertainty afecting
the recorded information. Sources of uncertainty include noise, human error, or limitations of
the information system supporting the process. Such imprecisions afecting the event data are
either recorded in an information system with the data itself or reconstructed in a subsequent
processing step, often with the aid of domain knowledge provided by process experts. Recently,
the possible types of uncertain data have been classified in a taxonomy, and efective process
mining algorithms for uncertain event data have been introduced [
The remainder of the paper is structured as follows. Section 2 formally describes uncertain event data. Section 3 introduces an extension to the XES standard capable of representing uncertain event data. Lastly, Section 4 concludes the paper.
In order to more clearly visualize the structure of the attributes in uncertain events, let us consider the following process instance, which is a simplified version of actually occurring anomalies, e.g., in the processes of the healthcare domain. An elderly patient enrolls in a clinical trial for an experimental treatment against myeloproliferative neoplasms, a class of blood cancers. This enrollment includes a lab exam and a visit with a specialist; then, the treatment can begin. The lab exam, performed on the 8th of July, finds a low level of platelets in the blood of the patient (event 2), a condition known as thrombocytopenia (TP). During the visit on the 10th of July, the patient reports an episode of night sweats on the night of the 5th of July, prior to the lab exam (event 1). The medic notes this but also hypothesizes that it might not be a symptom, since it can be caused either by the condition or by external factors (such as very warm weather). The medic also reads the medical records of the patient and sees that, shortly prior to the lab exam, the patient was undergoing a heparin treatment (a blood-thinning medication) to prevent blood clots. The thrombocytopenia, detected by the lab exam, can then be either primary (caused by the blood cancer) or secondary (caused by other factors, such as a concomitant condition). Finally, the medic finds an enlargement of the spleen (splenomegaly) in the patient (event 3). It is unclear when this condition has developed: it might have appeared at any moment prior to that point. These events are collected and recorded in the trace shown in Table 1 within the hospital’s information system.
In this trace, the rightmost column refers to event indeterminacy: in this case, 1 has been recorded, but it might not have occurred in reality, and is marked with a “?” symbol. Event 2 has more than one possible activity label, either PrTP or SecTP (primary or secondary thrombocytopenia, respectively). Lastly, event 3 has an uncertain timestamp, and might have happened at any point in time between the 4th and 10th of July. These uncertain attributes do not describe the probability of the possible outcomes, and we refer to such situation as strong uncertainty.
In some cases, uncertain events have probability values associated with them. In the example described above, suppose the medic estimates that there is a high chance (90%) that the thrombocytopenia is primary (caused by the cancer). Furthermore, if the splenomegaly is suspected to have developed three days prior to the visit, which takes place on the 10th of July, the timestamp of event 3 may be described through a Gaussian curve with = 7 . When probability is available, such attributes are afected by weak uncertainty.
Let us now describe a data standard extension able to represent strong and weak uncertainty, enabling the analysis of uncertain data with process science techniques.
The XES standard is designed to efectively represent and transfer event data, thanks to the descriptors extended from the XML language. Additionally, XES has been designed for flexibility: its descriptors, containers, and datatypes can be extended to define new types of information.
Figure 1 describes an extension of the XES standard able to represent uncertain data as described in the previous section and illustrated in the running example of Table 1.
This proposed extension can represent all scenarios of uncertain data shown in Section 2. As a consequence, it enables XES-compliant software to import and export uncertain event
value entry
0..n contains
value xs:any_datatype 0..n
contains
Log c o n t a i n s 0..n Trace tcno cno isan t a i n s 0..n Event c o n t a i sn 0..n Attribute 0..n contains
xs:double xs:double xs:double 0..n contains
list
2 orders
data, and it allows uncertainty-aware process mining tools to implement process discovery and conformance checking approaches on uncertain data, as described in the literature.
An example of a tool able to exploit this extended XES representation to manage and analyze
uncertain event data is the PROVED project1, which ofers process mining and data visualization
techniques capable of handling uncertain event data [
It is important however to emphasize the fact that the use of the extension described here is
not limited to the PROVED tool. There exist multiple tools able to support the XES standard,
such as ProM [
A set of synthetic uncertain event logs is publicly available for download2. In the same folder, it is possible to find the additional document (part of the BPM Resource track submission) explaining more in detail how our extension proposal models uncertain event data3.
Recent literature in the rapidly-growing field of process mining shows how descriptions of specific data anomalies can be extracted from information systems or obtained through domain knowledge. Anomalies labeled by such descriptions characterize uncertain event data, and there exist process mining algorithms able to exploit this meta-information to gain insights about the process with a precisely bounded reliability. A fundamental part of these data analysis approaches is however needed: formats for data representation and transmission. In this paper, we described an extension of the XES data standard which enables representation of such uncertain data, and that allows uncertain event to be read and written by existing XES-compliant software. This, in turn, empowers process mining researchers and practitioners to build analysis techniques that account for data uncertainty, and that can thus be more trustworthy and reliable.
We thank the Alexander von Humboldt (AvH) Stiftung for supporting our research interactions. We thank and acknowledge Fabian Rempfer for his valuable input on writing style, and Majid Rafiei for his contribution to the graphics.
1https://github.com/proved-py/ 2https://github.com/proved-py/proved-core/tree/An_XES_Extension_for_Uncertain_Event_Data/data 3https://github.com/proved-py/proved-core/blob/An_XES_Extension_for_Uncertain_Event_Data/data/ uncertainty_XES_standard.pdf