|The Application of Machine Learning to Improve ABA-Based Treatment Outcomes in Autism Spectrum Disorder|
|Sunday, May 27, 2018|
|4:00 PM–4:50 PM |
|Manchester Grand Hyatt, Seaport Ballroom F|
|Area: AUT/PRA; Domain: Applied Research|
|Chair: Erik Linstead (Chapman University)|
|Discussant: Erik Linstead (Chapman University)|
Researchers in applied behavior analysis (ABA) have yet to leverage "big data" on the same scale as other fields of research; however, as digital data collection systems become ubiquitous, and advancements in easy and affordable data analysis emerge, the application of machine learning techniques are becoming available to mainstream behavior analysts. Indeed, given the importance of collecting accurate and detailed data in ABA, the field is ripe for the application of machine learning approaches. While the treatment of Autism Spectrum Disorder is only one facet of ABA, there has been a notable increase in research literature evaluating the effectiveness of machine learning for diagnosing ASD, exploring its genetic underpinnings, and designing effective interventions. This symposium presents two applications of machine learning techniques to improve ABA-based treatment outcomes in ASD.
|Keyword(s): Autism, Data Mining, Machine Learning|
Identifying Profiles of Challenging Behaviors With Unsupervised Machine Learning
|Elizabeth Stevens (Chapman University), Esther Hong (Center for Autism and Related Disorders), Charly DeNocker (Center for Autism and Related Disorders), ERIK LINSTEAD (Chapman University)|
Individuals with autism spectrum disorder (ASD) are at a greater risk for challenging behavior than individuals with other developmental disabilities. An essential step in the treatment of these behaviors is the identification of the specific topography and function of the behavior. In the current study, data were collected from a large database, in which supervising clinicians recorded the topography and function(s) of behaviors treated as a part of an individual's behavior intervention plan. In a sample of 3,216 individuals with ASD, we report on the frequency of the most common challenging behaviors and the identified function of the behavior. We apply cluster analysis to a sample of 2,116 children with Autism Spectrum Disorder in order to identify patterns of challenging behaviors observed in home and center-based clinical settings. Results indicate that while the presence of multiple challenging behaviors is common, in most cases a dominant behavior emerges. This work provides a basis for future studies to understand the relationship of challenging behavior profiles to learning outcomes, with the ultimate goal of providing personalized therapeutic interventions with maximum efficacy and minimum time and cost.
Identification of Diverse Behavioral Phenotypes in Autism Spectrum Disorder
|DENNIS DIXON (Center for Autism and Related Disorders), Elizabeth Stevens (Chapman University), Marlena Novack (Center for Autism and Related Disorders), Esther Hong (Center for Autism and Related Disorders), Erik Linstead (Chapman University)|
Autism spectrum disorder (ASD) is a heterogeneous disorder, and individuals diagnosed with ASD exhibit diverse symptom presentation and severity, etiology, and treatment response. Research has explored potential ASD subgroups with preliminary evidence supporting the existence of behaviorally and genetically distinct subgroups; however, research has yet to examine treatment response across ASD subgroups. The present study investigated behavioral phenotypes in a large sample of children with ASD (N = 2400). Clustering methods were applied, revealing 16 subgroups. Further examination of the subgroups suggested 2 overlying behavioral phenotypes with a unique deficit profile each composed of subgroups that differed in severity of those deficits. Retrospective treatment data was available for a portion of the sample (n = 1,034). Treatment response was examined within each subgroup via linear regression methods. Results indicate that clustering has the effect of homogenizing treatment response, with over 70% of variance being explained by our models. These findings have implications on prognosis and targeted treatment of ASD.