A hybrid friend-based recommendation system using the combination of Meta-heuristic Invasive weed and genetic algorithms

One of the most important goals of researchers in designing recommender systems is to increase the accuracy of recommender models. The main purpose of this study is to combine weed algorithm and genetics to increase the efficiency of clustering, which increases the accuracy of clustering for analyzing suggestions in the recommender system. Clustering is based on three similarity criteria KMEAN, JACCARD and MINKOWSKI and in addition, it has been implemented with the mentioned algorithms. Using clustering method considering the mutation of genetic algorithm and using weed algorithm has led to the emergence of an efficient system that for this purpose, genetic generators have been used to sample data in the problem space instead of random generators, This selection has led to an increase in the accuracy of the algorithm due to the more uniform coverage of the problem space and the increase in the variety of problem searches. The Recommender system on the standard MovieLense data set is tested and its error is 0.02, which has a more minimal error than other algorithms (genetics and weeds).