Using Cogency and Machine Learning for Autism Detection from a Preliminary Symptom

Detecting the type of autism from a preliminary symptom in its early stage is difficult, especially in small children. This is because the symptoms of autism are based on the response of a child to cognitive functions. However if autism is not detected and treated between the ages of 20-60 months, the treatment at a later stage becomes more difficult. Additionally many parents are unable to express all the symptoms or behaviors which his/her child is experiences. In this paper we have proposed a system for finding all the possible symptoms from the preliminary symptom by using the previous long term patient records to diagnose the proper type of autism. This proposed technique applies both the machine learning algorithm and the confabulation theory to prepare a matrix which helps us to get all type of associations for generating rules. Confabulation theory is inspired by cogency. Cogency is used to compute the frequencies of occurrences on a pair of conditional items, hence we have used it for pairing of symptoms to predict the next most probable and confident symptoms for proper diagnosis of the type of autism. Using this technique we can identify both the common and rare types of autism. We have observed that our autism diagnosis tool uses less memory for its execution and it is faster because of one time database access as compared to using the Apriori algorithm.