Robotic Process Automation (RPA) is a technology that use computer programs called “bots” to mimic the sequence of actions taken by a human through a series of computer-based applications in order to accomplish a particular task. The sequence of actions that bots mimic is referred to as ``bot-flow’‘. A ``bot-flow’’ is typically comprised of various computer-human interactions, such as keyboard strokes, mouse clicks, and launching an application.
Currently, a design blueprint of a bot-flow is needed in order develop a bot. Such a blueprint is largely obtained through manual efforts by observing and mapping human actions into some process models. The reliant on manual efforts is not ideal as it is time-consuming and it may not be as comprehensive due to limited observation time window.
Fortunately, today’s IT systems have the capability to record user interaction data (i.e. data related to a human’s actions in his/her interactions with computer applications). This research project aims to develop algorithms and techniques that can be used to automatically discover a bot-flow from low-level user interaction data. A key challenge in this research is the vast volume of user interaction data which may contain a high level of random user activities that are not related to the main task being performed, as well as the interleaving of activities when users were multi-tasking.
Research activities in this project vary, depending on students’ skills and duration of the project. In general, research activities for this project include:
- Literature review on the state-of-the-art techniques that attempt to discover bot-flow from low-level user interaction data (VRES)
- The investigation of the applicability of various data mining techniques that can discriminate between task-relevant activities and non-task-relevant activities from user interaction log (VRES, Honours)
- Development of new data mining techniques to isolate task-relevant data (Masters, PhD)
- The development, implementation, and evaluation of algorithms to discover bot-flow from user interaction log (PhD)
The expected project outcomes are dependent on the scope of the project that students’ undertake. Key outcomes include:
- Gap analysis in the domain of bot-flow discovery
- Evaluation of the suitability of existing data mining techniques to discriminate, from user interaction log, relevant and non-relevant user activity with respect to a particular task
- New techniques that can isolate relevant user activities with respect to a particular task,
- Techniques that can semi-automatically discover bot-flows that are expressed as process models.
Skills and experience
- Familiarity with the fields of data mining, data science, process mining, process automation
- Reasonable writing skills
- Problem-solving and logical thinking capabilities
- Computer programming skills
Contact the supervisor for more information.