Aplicação de clustering e métricas à análise de LOG para avaliação automática de usabilidade de aplicações internet ricas

Abstract

The development of systems with effectiveness, efficiency and user satisfaction has made that usability become an important feature in quality assessment of software products. The analysis of user interaction data with the system is one way to measure this feature. New technologies allow the creation of systems with focus on quality of user interaction and a large part of the processing occurs on the client side, as Rich Internet Applications, from RIAs for short. However, user interaction data stored on the webserver are insufficient to extract useful knowledge about the quality of interaction. This dissertation presents an approach for automatic usability evaluation by applying clustering and metrics in log analysis of RIAs. The proposed approach uses the Web Application Usage Tracking Tool, WAUTT, to capture user interaction and the X-means algorithm to perform clustering. The result of the automatic evaluation applies metrics that provide quantitative information about the usability of web systems and may be used to assist evaluators in making decisions. Information obtained from metrics were compared to a traditional evaluation method, heuristic evaluation, to corroborate the results obtained by the proposed approach. Clustering was a useful technique to reduce the volume of data and allowed the evaluator focus their efforts on certain groups of users with similar behaviors. The application of metrics enabled perform a quantitative usability evaluation to assist the decision making process about some usability subcharacteristics. The automatic usability evaluation conducted as the proposed approach has the potential to assist in developing web systems.