Abstract
Nowadays, the transition of the customers’ desire to move forward from traditional to intelligent systems has caused rising trends of Ecommerce businesses. This can bring lots of opportunities and challenges for internet businesses to absorb the customer desire in a competitive manner. In this regard, recommender systems help users to find and select their desired items. These systems cannot recommend without having enough information about users and their desired items such as film, music, book. One of the main goals in these systems is to collect various information about user interest and available items of the system. Most of these systems operate based on collaborative filtering method in which similarity measures are used to select similar neighbors for a user and then the recommendation is offered based on the evaluation of their comments. In this paper, a recommender system using GMDH Neural Network algorithm is proposed to recommend films to users. The proposed model is based on exploring implicit trust from active users’ rates. Implicit trust networks among users are used to reduce prediction error of the improved user-oriented collaborative filtering algorithm. GMDH Neural Network offers a high learning speed even when there are few numbers of training data due to using the evolutionary genetic algorithm for the optimal design of the network structure. The proposed model is implemented using GevoM on MovieLens datasets and the results are compared with other algorithms in terms of Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). Results show that the proposed model outperforms other algorithms like MLP, Naïve Bayesian, J48, Bagging, SMO, RBF network, Logistic, and Random Forest with precision of 76% and absolute mean error of 0.273