Explorative Data Analysis of Eye-tracking Datafor Cognitive Assessments

This thesis evaluates a newly developed cognitive screening test using machine learning techniques. Existing cognitive tests can be challenging for individuals with multiple cognitive disabilities to complete; hence, this new test uses eye-tracking to facilitate performing the test. Analyzing the data collected from participants taking the test by building a pipeline, this thesis employs a Random Forest regressor to build a model and evaluates it using two existing cognitive tests: the Montreal Cognitive Assessment (MoCA) and the Mini-Mental State Examination (MMSE). The evaluation reports a reasonably good average prediction error value for both the MoCA and MMSE, which argues for an acceptable model. However, reports of a high variation among the prediction errors suggest that a further analysis of the test is advisable. Selection among features suggests that a feature set should include a variety of features, encompassing both spatial and temporal aspects, to enhance performance. While the eye-tracking test is currently under development, the thesis concludes that it may be of great use to health professionals in the future, aiding them in their work.