Data Mining for the Analysis of Eye Tracking Records

Abstract

It is proposed the implementation of a methodology for the analysis and classification of large volumes of records. It is studied and evaluated the application of DM as a tool to analysis qualitatively and quantitatively the register obtained by an eye movement tracking device, eye-tracking, when bring under to people with different levels of orthographic knowledge (OK: High, Medium and Low), in the face of two tasks; (i) detection of spelling error and (ii) in the detection of a simple character, in the brief exposure (1500 milliseconds) of words without and with misspelling. It used some analytical procedure series of DM such as: the search for response patterns; the creation of secondary variables; the use of classification of trees and grouping the data (k-means). New models were created as of the distance between the position of the spelling error and the position of the gaze of the participants. Differences in the visual attention were found between the participants; in the same way, it was observed that the misspelling influences the performance of the task (ii), diverting visual attention to spelling error, in the participants with High OK. It is concluded that the DM helps to find the particularities of eye movements from large volumes of data that generates eye-tracking, which cannot be analysed with simple procedures.