Although information systems support a wide range of recreational and social activities, the pattern of users’ behavior in such systems is not clear. We extend the process mining technology to work on common event logs that have no workflow cases reference, and name the new technology as behavior pattern mining. This automated analyzing approach can accelerate the discovery of collaboration patterns and help finding the law of collaboration.


As information technology increasingly supported a wide range of recreational and social activities, more and more people were able to connect online to benefit from the flexible virtual spaces. The computer and network supported systems enabled organizations and individuals to communicate, to cooperate, and to compete in a more flat environment. However, these systems changed the behavior customs heavily as those helped users’ daily work. To release the behavior pattern of computer supported work, is one of the key points to understand the modern working method, and produces recommendation to improve information systems, as well as indicates how to raise the work efficiency of people.

Most of scientists analyze behavior patterns through psychological, social, and organizational perspectives. From the year 2001 to 2004, NSF (National Science Foundation) funded SOC project (Science of Collaboratories) [1] to survey more than a hundred of collaborative projects in multiple fields, and aimed to define, abstract, and codify the broad underlying technical and social elements that lead to successful collaboratories. In detail, SOC project analyzed these projects from money, information and communication flow for institutions, IT resources and users. We notice that the money, information and communication condition in a project is the phenomenon and results cause by collaborators’ behavior, and the analysis through such three kinds of flow is an indirect method to discover behavior pattern. In this paper, we emphasize the direct way to do research on behavior pattern. Our plan is to mine the event logs that recorded by information systems, and to reveal the hidden relation of behavior in logs. This automated analyzing approach can accelerate the discovery of behavior pattern efficiently.

The automatic pattern discovery technology is cross-disciplinary among KDD (Knowledge Discovery in Data) and machine learning. The most well-known application area to discover pattern in logs is web usage mining, which is defined as applying data mining techniques to log interactions between users and a website [2]. Web usage mining tries to find out the relationship between independent operations and the successful clicking sequence for specific task, and uses the mined rule to classify particular customers and improve personalization of systems. Another hot pattern discovery technology is process mining, it tries to extract an explicit process model from event logs, i.e., to create a process model given a log with events such that the model is consistent with the observed dynamic behavior [3]. Compare to web usage mining, process mining support major workflow patterns in target model, which is efficient to express complicated relations for multi users who cooperate to accomplish a task. In addition, with the workflow pattern as its symbols, process mining supports the rediscovery of workflow model from workflow logs naturally, which is known as workflow mining. Further more, many researchers in process mining area use process mining as the same term of workflow mining. Although workflow mining doesn’t assume the presence of a workflow management system, it assumes that workflow logs with event data can be collected [4]. Such workflow logs are usually known as workflow traces.

In recent years, workflow mining technology developed a range of effective algorithms and approaches, and starts to apply on the real world. This paper discusses the problem to immigrate workflow mining technology on common system event logs but workflow traces, which will extend the application area of automatic pattern discovering method. We name the behavior pattern discovering on common system event logs as the term behavior pattern mining, and consider it as another sub field of process mining besides workflow mining. We discuss the requirements, challenges, structure of data source, possible approaches, and related tools of it basing on the achievement of workflow mining.


Behavior pattern mining is to discover the law of relationships and restricts between users’ behavior connected to a common goal from event logs, and targets to represent how people accomplish their works in multi users’ interaction environment. Behavior pattern mining is a sub field of process mining, which targets the automatic discovery of information from an event log and the discovered information can be used to deploy new systems that support the execution of business processes or as a feedback tool that helps in auditing, analyzing and improving already enacted business processes [5]. Workflow mining is another sub field of process mining, and workflow mining is the reverse engineering of workflow modeling basing on workflow logs.

Behavior pattern mining and workflow mining have some common points, such as both of them use workflow patterns as the symbols to present the mined results, and some mining approaches can be applied on both the two. On the contrary, the two are different in data source format and their target. Workflow mining grounds on workflow trace logs, whereas behavior pattern mining grounds on common event logs that are not partitioned. Behavior pattern mining focuses on how to structure and present users’ interactions, whereas workflow mining mainly focuses on workflow model improvement and verification. The knowledge from workflow mining area is a valuable base to discuss behavior pattern mining.

GlossyBlue theme adapted by David Gilbert
Powered by PmWiki