ABSTRACT
This chapter proposes an IoT-based auxiliary diagnosis method for autism spectrum disorder (ASD) that addresses the issue of label noise using label distribution learning (LDL), addresses the issue of sample imbalance using a cost-sensitive mechanism, and utilizes a support vector regression (SVR)–based method. By mapping the samples to the feature space, the label distribution learning approach overcomes the classification challenges imposed by high-dimensional features and eventually enables the IoT-based auxiliary diagnosis of multiclass ASD. The experimental findings demonstrate that, in comparison to existing approaches, the suggested method eliminates the imbalance between the majority and minority classes’ effect on the results. It successfully addresses the issue of unbalanced data in ASD diagnosis, provides more accurate and consistent classification performance, and aids in ASD diagnosis.
