Research On Recommendation Algorithm Based On Improved K-means Algorithm And SVD Algorithm

With the rapid development of the Internet,recommendation systems have entered the public view,and recommendation algorithms have emerged one after another.The movie recommendation algorithm is to provide users with an accurate and highly satisfactory movie recommendation list by modeling the user's personal information and historical operation behaviors,which can enable users and movie websites to achieve a win-win situation.For the problem of sparse movie data set,this article first populates the base data set with the average of each user's movie score,improve the recommendation accuracy,and uses the SVD singular value decomposition algorithm to reduce the dimension of the filled data set;Secondly,the elbow method and the improved K-means algorithm are adopted to determine the number of clusters K and the cluster center.In the traditional K-means clustering algorithm,the selection of initial cluster centers and k is very sensitive to the clustering results and will directly affect the clustering effect,so this article uses the elbow method to estimate the number of clusters k,and then uses the maximum and minimum distance algorithm to improve the K-means algorithm,completes the selection of the initial cluster center,and according to the distance between the user and the initial cluster center,update the average value of each cluster user as the cluster center,the final cluster and cluster center are obtained;Finally,the K nearest neighbor algorithm is used to calculate the similarity of the target user and classify it.When the target is found After the user belongs to the cluster,the target user's category is consistent with the category mark of the cluster to which the user belongs,and find the nearest neighbor with the target user in this category,and form a recommendation list based on these nearest neighbor high-rated movies,and recommend them to the user.A simulation experiment was carried out on the Movie Lens movie data set.The experimental results show that the algorithm in this paper reduces the scope of comparison,and at the same time,the accuracy and efficiency of recommendation and clustering are improved,realizes the diversification of movie recommendation,and improved the user's use Sense of experience.