ESOA:Classification of Autism Spectrum Disorder based on Gaze Tracking Imaging using Egret Swarm Optimization Algorithm

Autism spectrum disorder (ASD) may be appropriately identified using a set of cognitive tests, which is a time-consuming procedure, but early identification is usually difficult since people without ASD display comparable behaviors. To improve the ASD detection process, this paper presents a novel Egret Swarm Optimization Algorithm (ESOA) based on rough sets to detect ASD in patients by tracing the key features in their eye gaze. The noisy and irrelevant scan paths in the input high-dimensional eye gaze image are eliminated via the ESOA-based rough sets. The three steps in the proposed model are feature extraction, high dimensionality reduction, and classification. The accuracy of the Bayesian Network (BN) classifier is improved via the ESOA algorithm whose fitness is mainly based on the number of features selected and the rough set dependency degree. In this manner, the detection performance of the BN classifier is improved. The experimental results conducted on the eye tracking dataset show that the proposed model offers an improved accuracy nearly equal to 98%.