Optimized Active Fuzzy Deep Federated Learning for predicting autism spectrum disorder

Optimized Active Fuzzy Deep Federated Learning is a method based on fuzzy deep learning, federated learning, and active learning. This method has been developed for the classification and prediction of autism syndrome. Since there is no accurate medical test for autism, it is required to use machine learning for recognize it. Fuzzy deep learning is selected for the diagnosis of autism because the decision is unclear with the multiple questionnaires. The OAFDFL consists of three components, which include federated learning, deep fuzzy learning, and active learning. Federated learning is implemented to solve the problem of data scarcity. Deep fuzzy learning has been used for uncertainty in decision-making. Active learning has been used by psychologists to resolve the problem of data altering during the time. The main objective is developing Optimized Active Fuzzy Deep Federated Learning for predicting autism spectrum disorders with a long-term solution. Experimental results show that the proposed method outperforms other machine learning methods. According to the results, the best F-score, recall, and precision were obtained using the proposed method, as 90, 89, and 88%, respectively. Furthermore, the value of the ROC is equal to 0.905, and the Empiric ROC Area is equal to 0.892 for OAFDFL.