ASD stands for autism spectrum disorder which is a neurodevelopmental disorder that brings lots of health costs and has demonstrated troubles in communicating and participating in social interactions. The identification of ASD involves an assessment of a child’s development profiles as well as behavior patterns; nevertheless, the labeling that comes with the disorder makes children in question to be diagnosed only many years later. Some may not be so bold and when families label them as shy or hesitant, early diagnosis becomes a problem. The increasing rate of ASD is a good reason to focus on the development of the early diagnostic signs. The present work aims to apply feature selection procedures as machine learning approaches that can help improve diagnostic accuracy when compared to different feature selection techniques. To compare the working of the different algorithms, it was noted that gain ratio with rankers exhibited the highest level of accuracy among all the most popular algorithms. Although this paper covers several feature selection techniques, there is potential for more development of additional accurate predictive methods.