Detection of Web Service Refactoring Opportunities by

ii DEDICATION To My Family. iii ACKNOWLEDGEMENTS It is with a great joy that I reserve these few lines of gratitude and deep appreciation to all those who directly or indirectly contributed to the completion of this work: First and foremost I offer my sincerest gratitude to Dr. Marouane Kessentini, who dedicated all his wonderful time to collaborate, support and lead me to the end of this piece of work. His advices, dedication, availability, relevant comments, corrections and committeemen led to the success of this work. I also express my greatest thanks to SBSE members who supported me with valuable feedback and always kindly encouraged me to succeed this project. I thank all the lecturers of the CIS master degree who have used their valuable time to transmit the knowledge that help in putting this work together me. Finally, I wish to express my deep gratitude and thank my family who has consistently expressed its unconditional support and encouragement. All those who contributed in one way or another, to make this work, can be found here, the crowning of their efforts. Table 1: List of Web service interface metrics used. 14 Table 2: List of Web service code metrics used 14 Table 3: Web services used in the empirical study. ABSTRACT We propose, in this thesis, to consider the problem of Web service antipatterns detection as a multi-objective problem where examples of Web service antipatterns and well-designed code are used to generate detection rules. To this end, we use multi-objective genetic programming (MOGP) to find the best combination of metrics that maximizes the detection of Web service antipattern examples and minimizes the detection of well-designed Web service design examples. We report the results of an empirical study using 8 different types of common Web service antipatterns. We compared our multi-objective formulation with random search, one existing mono-objective approach, and one state-of-the-art detection technique not based on heuristic search. Statistical analysis of the obtained results demonstrates that our approach is efficient in antipattern detection, on average, with a precision score of 94% and a recall score of 92%.