Comparing Classification via Regression and Random Committee for Automatic Sleep Stage Classification in Autism Patients

Abstract

The prevalence of autism children has increased rapidly in the last few periods. There is no cure for autism. But the management and treatment of accompanying medical conditions can be done. One of the effects of his medical condition is a sleep disorder. But children with autism have difficulty communicating the disorders they experience. In medicine, the detection of sleep disorders can be done through a test called polysomnography. One of the purposes of this test is to find the patient's sleep patterns through the sleep stage classification. But the doctors need several days to analyze each test. This study proposes an application that can classify it automatically. The method used was based on machine learning. The two classifiers were classification via regression and random committee. The both performances were compared in sleep stages classification for the autism patients. The result showed that random committees had a higher performance than classification via regression. Its performance got more than 85% for accuracy, precision, recall, and F-measure. This study also implemented resampling to overcome imbalance class problems. It can be seen that this process was useful in improving the performance of both classifiers.