Music personalized recommendation system based on improved KNN algorithm

The rapid development of the Internet has led to the emergence of massive amounts of music. In order to make people hear their favorite music more accurately, various recommendation algorithms emerge one after another. KNN is one of the best neighboring algorithms based on collaborative filtering. However, when the ratings that from the users' changed, the error rate of the KNN recommendation algorithm is higher. In this paper, we point the disadvantages of the KNN algorithm that is greatly affected by the rating. And then we propose the improved algorithm KNN-Improved which is based on the KNN algorithm and makes use of the mean value's thought of the Baseline algorithm, besides, we add to the standard deviation of the rating. These measures can effectively reduce the impact of too high or too low about user ratings, reduce the error rate of the recommendation algorithm, and then achieve better recommendation results. Finally, different recommendation algorithms are compared for performance and improved algorithms are applied.