With the growing data on the Internet, recommender systems have been able to predict users’ preferences and offer related movies. Collaborative filtering is one of the most popular algorithms in these systems. The main purpose of collaborative filtering is to find the users or the same items using the rating matrix. By increasing the number of users and items, this algorithm suffers from the scalability problem. On the other hand, due to the unavailability of a large number of user preferences for different items, there is a cold start problem for a new user or item that has a significant impact on system performance. The purpose of this paper is to design a movie recommender system named TRSM-RS using users’ demographic information (just users’ gender) along with the new weighted similarity measure. By segmenting users based on their gender, the scalability problem is improved and by considering the reliability of the users’ similarity as the weight in the new similarity measure (Tanimoto Reliability Similarity Measure, TRSM), the effect of the cold-start problem is undermined and the performance of the system is improved. Experiments were performed on the MovieLens dataset and the system was evaluated using mean absolute error (MAE), Accuracy, Precision and Recall metrics. The results of the experiments indicate improved performance (accuracy and precision) and system error rate compared to other research methods of the researchers. The maximum improved MAE rate of the system for men and women is 5.5% and 13.8%, respectively.