Data Science Techniques for Modelling Execution Tracing Quality

This research presents how to handle a research problem when the research variables are still unknown, and no quantitative study is possible; how to identify the research variables, to be able to perform a quantitative research, how to collect data by means of the research variables identified, and how to carry out modelling with the considerations of the specificities of the problem domain. In addition, validation is also encompassed in the scope of modelling in the current study. Thus, the work presented in this thesis comprises the typical stages a complex data science problem requires, including qualitative and quantitative research, data collection, modelling of vagueness and uncertainty, and the leverage of artificial intelligence to gain such insights, which are impossible with traditional methods. The problem domain of the research conducted encompasses software product quality modelling, and assessment, with particular focus on execution tracing quality. The terms execution tracing quality and logging are used interchangeably throughout the thesis. The research methods and mathematical tools used allow considering uncertainty and vagueness inherently associated with the quality measurement and assessment process through which reality can be approximated more appropriately in comparison to plain statistical modelling techniques. Furthermore, the modelling approach offers direct insights into the problem domain by the application of linguistic rules, which is an additional advantage. The thesis reports (1) an in-depth investigation of all the identified software product quality models, (2) a unified summary of the identified software product quality models with their terminologies and concepts, (3) the identification of the variables influencing execution tracing quality, (4) the quality model constructed to describe execution tracing quality, and (5) the link of the constructed quality model to the quality model of the ISO/IEC 25010 standard, with the possibility of tailoring to specific project needs. Further work, outside the frames of this PhD thesis, would also be useful as presented in the study: (1) to define application-project profiles to assist tailoring the quality model for execution tracing to specific application and project domains, and (2) to approximate the present quality model for execution tracing, within defined bounds, by simpler mathematical approaches. In conclusion, the research contributes to (1) supporting the daily work of software professionals, who need to analyse execution traces; (2) raising awareness that execution tracing quality has a huge impact on software development, software maintenance and on the professionals involved in the different stages of the software development life-cycle; (3) providing a framework in which the present endeavours for log improvements can be placed, and (4) suggesting an extension of the ISO/IEC 25010 standard by linking the constructed quality model to that. In addition, in the scope of the qualitative research methodology, the current PhD thesis contributes to the knowledge of research methods with determining a saturation point in the course of the data collection process.