Recommender systems have been introduced to provide a solution to navigating the huge volume of information already available and growing at an explosive rate. The amount of information available in electronic form, such as news, movies, books, advertisements and other online information is overwhelming us. Recommender systems are computerbased techniques that can be utilized to efficiently provide personalized services in many e-business domains. In this thesis, recommender system has been designed by mixing two main types of recommender systems (content based on personal profile and collaborative based). This type of system producing recommendations for its users in two stages. In the first stage, searching about active user's neighborhood is done to compute the similarity with the active user. The similarity is computed in two steps, the first step is to compute personal similarity using content based technique, depending on the personal features only. The second step is a conditional step that is if the user has enough rating then the similarity computed using collaborative filtering technique depending on the user ratings (rating similarity) in addition to personal similarity computed by the first step. In the second stage a list of new items is recommended from highly rated items by nearest neighbor users, with or without predictions on the acceptance of the list by the user. The content based part in which a personal similarity is computed a weight for each personal feature is required. So in this work, a survey has been made to obtain initial value for impact ratio (weight) for the effectiveness of each feature. Then the computation of these ratios is updated from time to time according to the given new users information. These updates are made according to Mean Absolute Error (MAE) between the real ratings and prediction of ratings.