LEARNING MODELS FOR QUALITY ASSESSMENT IN WEB-BASED SOFTWARE

Abstract

Software quality is one of the pivotal aspects of the software development industry which ensures product compliance to the requirement specification and standards. Conventional software development was mostly related with building desktop applications. The past decade has seen a proliferation of architectures, frameworks, and languages in software development. Software methodologies have shifted from building monolithic standalone applications to service-oriented, metric-driven, collaborative agile-based development of Web-based software. Web analytics is the process of examining websites to uncover patterns, correlations, trends, insights and other useful information which can be utilized to optimize web usage and to improve the quality of website. A Website quality model essentially consists of a set of criteria used to determine if a website reaches certain levels of fineness. UX (or user experience) directly measures the quality of site interactions, and is an indirect representative of site success and customer conversions. That is, a bad UX bounces away visitors to seek a more reliable website. Every single second a user spends on a website is directly attributable to the usability of a good UX. Hence, the evaluation of quality of websites is essential to determine user acceptance, that is, the users are the parameter measured for the success of the site. The work presented in this research expounds the evident shift of quality models for conventional software to web-based software. It further suggests a π-model representation for quality criterion relationship interpretation for both types of software. The horizontal line of the π signifies the backbone of quality models with quality assessment parameters common to both kind of software whereas the two vertical pillars of the π depict the quality attributes specific to the software type. This research also proffers an approach which associates the website assessment with the user satisfaction and acceptance. The proposed WQA (Website Quality Analytic) Model considers websites from seven domains, namely, .com, .net, .org, .int, .gov, .edu and .mil and using 13 UX- based quality attributes evaluates the quality of websites in each domain. The quality assessment is automated using supervised learning models to predict good, average and bad websites. This feature (attribute) - based predictive model for quality analytics is empirically analyzed for five classification algorithms. A qualitative analysis of the domainwise classification of websites is presented too.