Abstract
Now that technology is developing rapidly, the field of computing has been showing exponential progress, and scientific and technological civilization of mankind is making continuous progress. Games, live broadcasts, and short videos are all new products under big data. Especially for short videos, the huge database records the preferences of billions of users. With the combination of collaborative filtering algorithms and content-based recommendation algorithms, computers can always accurately recommend suitable videos to various users. In this paper, improvement of this method is aimed at the item-based algorithm in the collaborative filtering algorithm. For the item-attribute matrix, first, use the Jaccard distance to calculate the similarity, and then use this similarity value instead of the Euler distance formula to bring it into the k-means clustering, and use iteration to obtain countless different clusters. Finally, set a threshold x, which is the distance between each cluster center. Whenever there is a new matrix to be classified, the similarity y corresponding to this matrix is calculated first. If y<x, the matrix is classified into the corresponding cluster. This approach can improve the diversity of recommended videos and tap the potential interests of users. Such improvements to the matrix can improve the accuracy of the algorithm and user stickiness.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call