Cross-Browser and Cross-Plataform incompatibilities automatic detection using machine learning in e-commerce

Abstract

Context: The e-commerce consumers can interact with the virtual stores through a variety the Web browsers, such as, Internet Explorer, Google Chrome, Mozilla Firefox, Apple Safari and others. The web browsers are executed in several operacional systems (Android, iOS, Windows, MacOS and Linux) and its installed in plataforms (desktops, smartphones and tablets). This flexibility that characterizes the portability of the e-commerces, on the other hand, it has conducted to the appearance of incompatibilities in the behavior and renderization of the layout of the e-commerces. It is expected that there will be standardized and uniform behavior independent of the regardless of how the user accesses the e-commerce. Recent advances in technology have raised the complexity of Web application development (e-commerce, institutional websites, internet banking). Although there are specifications that aim to ensure the homogeneity of the envirements, through of the use of strategies of the Design Responsive and frameworks of front-end (Bootstrap, Foundation and Materialize), the developers need to implement many customized versions of your web applications. Moreover, at the same time they need to keep an consistent set of features and layout between all the versions [Choudhary2014Xpert], therefore, an expensive task and prone the mistakes [Walsh et al. 2017].The observed differences in the layout or in the behavior of a web application when rendered in browser different have been referenced as Cross-browser Incompatibilities (XBIs) [Choudhary2010Webdiff2, Choudhary2012Crosscheck]. To overcome this problem during the software development process, developers and testers have the manual task of finding and to fixing the incompatibilities before making the e-commerce available to consumers.Goal: This research project aims to improve the approach of Cross-Browser and Cross-Plataform incompatibilities automatic detection through the use of machine learning (Artificial Intelligence).Method: The proposed approach segments a web application into multiple DOM (Document Object Model) elements. The detection task of XBI was modeled as a classification problem of machine learning using the following properties to compose the feature set: position differences, size differences, and image comparasion of each DOM element of the web application. A separate three-step experiment (data acquisition, manual classification and classification model evaluation) will be performed to investigate the effectiveness of the classification model. Expected Results: It is expected that the use of machine learning will reduce the amount of false positives presented by the state of art in this research topic and increase the effectiveness of the approach to enable the development of a web tool of Cross-browser and Cross-plataform incompatibilities automatic detection for e-commerce. Contribution: The viability of a web tool available as SAAS (Software as a Service) will provide for the company business the possibility of a globally scalable service. For the market, the tool will contribute to increase of the quality of the e-commerce and reduce the manual test costs to identify XBIs. Keywords: Cross-plataform Incompatibilities, Cross-browser Incompatibilities, Machine Learning, Software Testing, e-commerce. (AU)