University of Twente Student Theses

Login

Using user workflow analysis to create insights in content-intensive applications : Combining Process Mining and Model-Driven Engineering to create a reusable, scalable and user-friendly solution.

Hol, M.C. (2018) Using user workflow analysis to create insights in content-intensive applications : Combining Process Mining and Model-Driven Engineering to create a reusable, scalable and user-friendly solution.

[img] PDF
2MB
Abstract:Traditional methods for evaluating software are often less suitable for content-intensive applications because they are not always able to track the steps of a user across the entire application, because they are less suitable for continuous improvement due to their lack of scalability, and because the observed behavior might be different from reality. We introduce user workflow analysis as a solution to help content and software developers (user workflow analyzers) of content-intensive applications (referred to as System Under Analysis or SUA) to gain additional insights in the usage of the application that can’t be acquired through traditional methods. We present a design and prototype implementation of the User Workflow Analysis Tool (UWAT). This tool successfully implements user workflow analysis by combining the strength of Process Mining (PM) in extracting patterns from event log data and the strength of Model- Driven Engineering (MDE) in standardized and automatic transformations. This research shows that MDE, due to its standardization power, can mitigate the shortcoming of PM that it is often left up to user to configure the plugins and interpret the results. In this design, only the implementer of the UWAT needs to understand the PM configuration and results, and all other user workflow analyzers can benefit from this in a standardized, automated and userfriendly way. No references were found that this combination was used for this purpose before. The UWAT itself consists of five metamodels: SUA metamodel, plugin specification metamodel, user specification metamodel, Process Mining metamodel, and result metamodel. All metamodels are set-up as generic as possible, so they can easily be used for multiple SUAs. Four transformations are introduced: to create the SUA model instance, to translate the input of the user workflow analyzer to a user specification model instance and the instructions for PM, to process the result of PM to useful results for the user workflow analyzers based on the user specification, and finally to translate the result model instance to the output data that can be used for visualization of the results. The PM execution happens externally, but the UWAT handles the call to these external plugins. In this way, users of the UWAT can benefit from external knowledge in an area that is rapidly evolving. The visualization for the user workflow analyzers is separated from the UWAT, so the SUA implementer has a lot of flexibility in how to present the results. For the prototype implementation and validation, a case study is done for a content-intensive application, Mobina. The Mobina team created their own modelling technique for the complex content. They are expected to benefit from user workflow analysis to validate the modelling technique and the interface supporting this. The design was validated using two types of expert opinion: user validation and technical review. Potential user workflow analyzers were interviewed to identify the added value of user workflow analysis and to assess the requirements relevant for user workflow analyzers. The technical review is a critical review of this design, and where necessary its implementation, to validate to what extent the requirements of the architecture were fulfilled. This thesis shows that using user workflow analysis can provide a lot of added value to SUAs. By using the UWAT, the user workflow analyzers get more insights in the behavior of the users in the SUA and the effects of their work on their users. Eventually, this will lead to a better user experience and satisfied customers. Simple examples already have added value, but this could be even more with analyses that focus on the structure of the content and the software. This research successfully showed how by combining PM and MDE, different user workflow analyzers can benefit from analyses in a reusable, scalable and user-friendly way. This combination of PM and MDE led to a user-friendly solution where the user workflow analyzers are not bothered with the implementation details. New functionality and analyses can easily be added which makes the solution flexible and scalable. Due to the generic setup of the metamodels, most of the metamodels can be reused for different SUAs. Some parts of the transformations need to be implemented for each SUA separately, so the SUA implementer has a lot of flexibility and can implement all desired analyses. As soon as the necessary PM support is there, this solution can also be fully automated. User workflow analysis as presented shows a lot of potential, but we expect this can be even more. It will be interesting to see whether other techniques, e.g., machine learning, can improve or extend the user workflow analyses. Other types of applications than contentintensive applications could also benefit from user workflow analysis and this design, so it is useful to research this further. Finally, it will be very interesting to bring this to the next step of the design cycle. The design presented here contains all the core aspects, but there are also several remaining opportunities. There are possibilities to standardization and share one UWAT implementation amongst SUAs. When researching these possibilities more indepth, an even better support for user workflow analysis can be established.
Item Type:Essay (Master)
Clients:
Mobina IT B.V., Enschede, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:http://purl.utwente.nl/essays/76894
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page