1 | Page The Application of a Semantic-Based Process Mining Framework on a Learning Process Domain Kingsley Okoye a* , Syed Islam a , Usman Naeem a , Mhd Saeed Sharif a , Muhammad Awais Azam b and Armin Karami a a School of Architecture Computing and Engineering; College of Arts, Technologies and Innovation, University of East London, Docklands Campus, 4-6 University Way, London, United Kingdom. E16 2RD. {K.Okoye, Syed.Islam, U.Naeem, S.Sharif, A.Karami}@uel.ac.uk b Faculty of Telecom and Information Engineering, University of Engineering and Technology, Taxila, Pakistan. awais.azam@uettaxila.edu.pk Abstract — The process mining (PM) field combines techniques from computational intelligence which has been lately considered to encompass artificial intelligence (AI) or even the latter, augmented intelligence (AIs) systems, and the data mining (DM) to process modelling in order to analyze event logs. To this end, this paper presents a semantic-based process mining framework (SPMaAF) that exhibits high level of accuracy and conceptual reasoning capabilities particularly with its application in real world settings. The proposed framework proves useful towards the extraction, semantic preparation, and transformation of events log from any domain process into minable executable formats – with focus on supporting the further process of discovering, monitoring and improvement of the extracted processes through semantic-based analysis of the discovered models. Practically, the implementation of the proposed framework demonstrates the main contribution of this paper; as it presents a Semantic-Fuzzy mining approach that makes use of labels (i.e. concepts) within event logs about a domain process using a case study of the Learning Process. The paper provides a method which aims to allow for mining and improved analysis of the resulting process models through semantic – labelling (annotation), representation (ontology) and reasoning (reasoner). Consequently, the series of experimentations and semantically motivated algorithms shows that the proposed framework and its main application in real-world has the capacity of enhancing the PM results or outcomes from the syntactic to a much more abstraction levels. Keywords—process mining, process models, ontology, semantic annotation, reasoner, AI, event logs I. INTRODUCTION The need for novel approaches in design and integration of computational intelligence into everyday (e.g. business, learning) processes has sprout research investigations on how to exploit such tools for use in improving the ever increasing data about various organizations. In recent years, a common challenge with many of the business processes has been on how to develop intelligent systems that can provide platforms for exploring the additional, and most often, the monotonous tasks of managing the entire operational process and quality of information - by providing understandable and useful insights on the best possible ways to make the envisioned information explicable in reality. Such process-related analysis, often allied to process mining, means there is also need for tools and techniques that can extract valuable information from the event logs about the domain processes in view. Most organizations have devoted a greater amount of resources towards modelling of their everyday processes or operations. Nevertheless, majority of the mapped processes appears to be non-operational, unfitting and/or presents itself in formats that are targeted towards understanding the process flows (workflows) rather than tackling the complexity of the business or domain processes in reality. In other words, a greater number of the resulting models and methods tends to support just machine-readable systems rather than machine- understandable systems at large. By machine-understandable systems we refer to methods that are developed not just for representing information in formats that can be easily understood by humans, but also for creating applications and/or systems that trails to inclusively process the information that they contain or supports. Thus, the mapped processes (models) are either semantically-labelled (semantic annotation) to ease the analysis process, or represented in a formal structure (ontology) that permits a computer (through semantic reasoning) to deduce or discover new facts/knowledge as a result of the defined relations or assertions within the ontology. According to the work of [1] - [4] the efficiency or quality of the so-called organizations process can be improved if the process analysts and/or owners could perform an accurate analysis or exploration of the extracted event logs. For instance, the process of visualizing the relations or attributes the process instances share amongst themselves in a knowledge base. Of late, the PM field [3] or better still the latter - Process Querying [5] have turn out to be such valuable method that can be used to discover meaningful information from the recorded event logs. Besides, the process mining (PM) field combines techniques from computational intelligence and data mining to process modelling and analysis, as well as several other disciplines related to the business process management (BPM) [3] and AI to analyze the data. However, [6] and [2] notes that a common problem with majority of the PM methods in current literature is that they tend to rely on the tags or labels within the events log to discover the process models, and as consequence, appears to be vague and limited to some extent when confronted with unstructured data. The aforementioned challenge is as a result of lack of abstraction level of analysis (i.e. conceptual information)