Deep learning-based privacy-preserving recommendations in federated learning
Chandra Sekhar Kolli, V. V. Krishna Reddy, Tatireddy Subba Reddy, Mohan Kumar Chandol, Durga Bhavani Dasari, Mule RamaKrishna Reddy
Abstract
Privacy preservation in recommendations has been increasingly garnering huge interest from the research community owing to the rapid rise in data security and privacy concerns among users. The computation overhead and attaining high recommendation accuracy remain the key issues in the existing methods. In this research, a course recommendation method using Federated Learning (FL) based on Deep Learning is presented. The course recommendation technique is carried out in the local nodes using multiple phases, like agglomerative matrix generation, course grouping, bi-level matching, retrieval of learner-preferred courses, and course recommendation. Here, course grouping is accomplished using Deep Fuzzy Clustering (DFC), and Deep Convolutional Neural Networks (DCNN) performs recommendation. The DFC-DCNN-FL is examined based on accuracy, False Positive Rate (FPR), loss function, Mean Square Error (MSE), Root MSE (RMSE), and Mean Average Precision (MAP) and is found to have attained values of 0.909, 0.116, 0.126, 0.291, 0.539, and 0.925.