Detection of Autism Spectrum Disorder Effectively Using Modified Regression Algorithm

Person with impairments in social communication, abnormal behavior, and sensory activities are considered as suffering with neurodevelopmental syndrome and this syndrome is termed as autism spectrum disorder (ASD). Diagnosis process of ASD is based on observation of frequent movements, social communication skills. and eye contact of person. In some cases, standard questionnaires are used to assess the person. The objective of this paper is to automate the diagnosis process to generate accurate results, which are used to detect ASD. Diagnosing and predicting autism at early age help to take better treatment. Early detection of ASD in children can reduce the symptoms of ASD and they can mingle with normal children. In recent years, more research work has been done on ASD to find methodologies for ASD prediction. Machine-learning methods are efficient to analyze ASD as it generates accurate results with less computational power compared to other methods. An efficient regression-based algorithm is proposed to predict ASD with less computation time makes detection process faster. The proposed algorithm is applied over dataset collected from UCI machine-learning repository. Dataset consists of around 2000 people's information of varying age groups like adolescent, adult, child, and toddlers. The results obtained from this dataset are analyzed and present efficiency of algorithm to predict ASD at an early age. Performance analysis on proposed and existing algorithms is compared and analyzed.