For those coming across this term for the first time, ensemble learning is a meta-algorithm that combines various algorithms to create more stable predictions. This approach has already been applied in various fields, such as finance, marketing, and autism research.

Speaking of Autism Research, a group of researchers took on the challenge to leverage ensemble learning for the purpose of predicting treatment outcomes in individuals with autism. The benefits are quick time used for diagnosis and decision-making, reduced costs, and ultimately saving lives.

The Research Procedure

The procedure involved using five single classifiers, which were compared to several MCLS algorithms. In other words,  the researchers compared individual machine learning algorithms to a set of combined algorithms known as MCLS. This was done in order to determine which approach would result in better predictions for ASD Therapy.

Specifically, the study incorporated a dataset consisting of behavioral data and robot-enhanced treatment (intervention) versus standard human treatment (control). This dataset was collected from 3,000 sessions and over 300 hours of therapy conducted with 61 autistic children above the age of three.

The results of the treatment efficacy on autistic children was assessed by analyzing the changes between their initial and final ADOS scores. ADOS stands for Autism Diagnostic Observation Schedule, and it is a  standardized assessment tool used to diagnose and evaluate the severity of autism spectrum disorders.

The Result

Results showed that  when it comes to predicting ASDT, among single classifiers, decision trees are the best choice. This is supported by the results of this study, which showed a low 36% error rate for decision tree classifiers.

However, MCLs proved to be even more effective, as they combine the strengths of multiple classifiers, averaging out predictions from different models to come up with a more accurate result.

Research Implication

Bottomline, the findings of this study suggest that MCLs, best known as Machine Learning Classifiers, are a promising tool for predicting Autism Spectrum Disorder (ASD). Specifically, the use of static parallel MCLs with three classifiers, which include decision trees, k-nearest neighbor, and logistic discrimination, showed the highest effectiveness in predicting ASD.

However, it was also observed that factors such as eye contact and social interaction have a significant impact on the success of ASD-enhanced treatments compared to stereotypes, non-verbal speech, and social touch. It would be interesting to conduct future studies that compare autistic infants to autistic adults.

This could provide more insight into the progression of ASD understanding and how different cognitive systems may play a role in its onset. Parents seeking quality Treatment and Care for their child with ASD will be relieved to know that Illinois offers professional treatment, ABA  therapy, and quality care services at the Nevada Autism Center.