Autism Spectrum Disorder (ASD), a form of neurological sickness, appears in youngsters among 6 and 17 years of age and impacts conversation abilities and social conduct. ASD impacts social interactions and communication and reasons repetitive behaviors in patients. Consistent with the WHO, ASD impacts one baby in one hundred sixty. Early diagnosis during adolescence is critical and might improve the social capabilities and communiqué problems of youngsters. So, it is very important to find a proper way to detect Autism. For diagnosing neurological diseases, it is highly required to study the structural vicinity relationships within the brain. Hence, to study the brain and its patterns, fMRI (functional magnetic resonance imaging) is used. It detects correlated fluctuations inside the blood oxygen degree-dependent (bold) alerts from the brain regions. This paper is proposed with the primary motive of building, training and testing a 3D Convolutional Neural Network (CNN) model using a dataset abbreviated as ABIDE—Autism Brain Imaging Data Exchange dataset and predict outcome of the class labels as autism or no-autism for unseen health records. The project also aims to present a comprehensive analysis and provide some recommendations related to ASD results generated by the system.