Date

2021-7-14

Author

Khalaji, Erfan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

456
views

768
downloads

Cite This

Autism Spectrum Disorder (ASD) is a heterogeneous neurodevelopmental disorder that causes social, communication and behavioral challenges with different severity levels. Studies report a considerable increase in ASD prevalence during the past two decades, and clinical psychologists face difficulties identifying individuals with ASD. Researchers have been using different techniques such as eye-tracking to help address ASD diagnosis. A previous study shows that training a logistic regression model with eye-tracking data gathered using web pages can be an effective method to identify high-functioning autism in adults. This thesis uses the same eye-tracking datasets, each of which includes two web-related tasks. The first dataset includes a searching task and a freely browsing task, while the second dataset includes a synthesis task and a time-restricted browsing task. Our study investigates a different data preprocessing method and evaluates various machine learning models based on that. This study obtains an accuracy of 91.6% and 76.3% for the searching task and the freely-browsing task, respectively. In contrast, the previous study obtains an accuracy of 75% for the search task and 71% for the browse task. Our study and the previous study obtain lower accuracies using the second dataset. Therefore, the results in this thesis demonstrate that tasks and pages used can affect the accuracy of machine learning algorithms. This thesis also suggests eye-tracking on the web can be a complementary practical tool for human experts to diagnose people with ASD more precisely.

Subject Keywords

Autism Spectrum Disorder, Machine Learning, Eye-Tracking, Web

URI

https://hdl.handle.net/11511/91565

Collections

Northern Cyprus Campus, Thesis