Discriminant analysis for detection of low usability web pages

The purpose of this work is to reduce the cost of the web usability evaluation by usability testing. The cost will reduce by detecting low usability web pages. We analyzed empirically to find detectable metrics from the quantitative data including eye movement. We investigate the relation between the quantitative data about the behavior of users and the web usability evaluated by subjects. The subjects examined 263 web pages in total. The Logistic Regression Analysis (LRA) is used to detect the low usability web pages. The metric is defined using the browsing time, the amount of wheel rolling and the moving distance of gazing points which are selected by the stepwise selection method based on the Area Under the ROC Curve (AUC). When setting the cut-off value of 20 pages with low usability, the highest detection rate of the web pages with low usability is 0.65 by LRA. From the results, if the moving speed of gazing points is high, the possibility of low usability web page is high.