ARTIFICIAL NEURAL NETWORKS AS AN ALTERNATIVE AND COMPLEMENTARY TOOL TO HELP IN AUTISM DIAGNOSIS

• DOI: 10.22533/at.ed.9072322123

• Palavras-chave: -

• Keywords: Artificial Neural Networks. Learning algorithm. Computational models. Autism.

• Abstract: Artificial Neural Networks (ANNs) are computational models based on human brain structures. The ANNs receive data, learn from it and generate coherent responses to unknown data. Autism Spectrum Disorder (ASD), commonly known only as autism, has already been the subject of studies in the area of artificial intelligence to aid in the diagnosis, which, in turn, is very complicated because the symptoms appear from the age of two and are generally very light. This disorder is very complex and therefore it is not always possible to make an accurate diagnosis. Therefore, this work check the possibility of using of ANNs as an alternative and complementary tool to aid in the diagnosis of autism, due to their ability to learn from a reduced amount of data and generate coherent responses to new data. Tests were made using the MATLAB software with the nnstart and the pattern recognition application tool. For the first tests with answers from only 15 real people, 93.3% of correct answers were obtained with ANN. With the addition of fictitious responses totaling 150 responses, it was possible to reach 100% ANN accuracy.