Cost-Effective and Efficient Detection of Autism from Screening Test Data Using Light Gradient Boosting Machine

Abstract

Autism spectrum disorder (ASD) is a developmental disorder that affects the brain. Autism constrains a person’s ability to interact and communicate with others. The cause of autism, in general, is unknown though genetics does play a role in the manifestation of the condition. In the absence of clear identifiable biomarkers, shortcomings of the available prognostic approaches create a need for a new technique that is speedy, cost-efficient, and provides an error-free diagnosis. The system should also be able to adapt to the varying characteristics of subjects with ASD. The amelioration machine learning brings to automated medical diagnosis which has inspired us to come up with a solution. An adept screening and diagnostic test for patients exhibiting known autistic symptoms is a well-compiled, specific, and approved questionnaire, which facilitates an easy and cheap diagnosis. Autistic Spectrum Disorder Screening Test data is collected from one such questionnaire. We used a combination of three publicly available datasets containing records related to ASD in children, adolescents, and adults. There are a total of 1100 instances along with 21 attributes. The proposed study uses a Light Gradient Boost (LGB) based model for classification, along with Random Search for hyperparameter optimization, which yielded a high accuracy of 95.82%.