A Survey on Usability Evaluation using Traditional Methods and Machine Learning Algorithms in Last One Decade (2014-2024)

Usability constitutes a fundamental element that significantly contributes to the designing and development of successful interactive software applications. The users tend to prefer applications that effectively help to achieve the objectives with ease and satisfaction which will determine the success or failure. The choice of the most accurate method for a particular scenario is a very difficult decision, due to the existence of various approaches, described in the literature. Therefore, a survey was carried out which offers a comprehensive examination of usability evaluation methodologies (UEMs), effectively linking traditional practices with innovative machine learning (ML) based techniques. By scrutinizing data derived from 40 research studies, the survey addresses four principal research inquiries, delving into definitions of usability, frequently addressed usability attributes, established UEMs, and the application of ML techniques in usability evaluation. The results underscore that the International Standard Organization (ISO) 9241-11 definition is extensively adopted, with 20 of the studies acknowledging its importance. Attributes such as efficiency, satisfaction followed by learnability emerge as the predominant usability characteristics. Among the UEMs, questionnaires followed by surveys represent the most prevalently utilized methodologies. In the realm of ML, Support Vector Machines (SVM) and random forest are identified as the most frequently employed technique, followed by decision trees, Naïve Bayes, logistic regression, regression model and neural networks. These methodologies illustrate the potential of ML to augment usability evaluation thereby highlighting the necessity for research across varied domains to establish optimal practices and it is currently in a state of evolution.