M.A. Thesis:

Designing an advisory system Using Neural-Fuzzy Networks

By increasing development of the web and its various uses, access to the information users need has become an important issue. Therefore, it is necessary to have a system that intelligently recognizes the interests of users and offers them the nearest and most appropriate information in a short time. The purpose of web personalization is to discover the behavioral patterns of users interacting with the web and predict their future behaviors to achieve the desired goal. In this thesis, according to the preferences of the users of their favorite films, it is recommended. For this purpose, a reference method is first introduced whereby users' views of the films are used in a two-dimensional matrix. This two-dimensional matrix is then converted into two matrices that can be used to estimate the user's interest in each film. This method has a very low speed despite its high performance. Therefore, it has been attempted to increase the speed of the proposed method significantly by maintaining the performance of the reference method. The neural network is used for this purpose. The neural network is a method that teaches how to communicate between data and uses it for similar purposes. In this thesis, the perceptron neural network with the number of inputs and the hidden layers is specified. The main advantage of the proposed method is that of the algorithm speed reference and response time. In the reference method, the training process must be repeated at each step to offer the user, but only once in the proposed method. The results show that the proposed method having much similarity to the results of the reference method has a much higher speed than this method.

Keywords: web personalization, forecasting, neural Network, hidden Layer