Efficient Barrage Video Recommendation Algorithm Based on Convolutional and Recursive Neural Network

Abstract

Convolutional neural network has been widely used in recommendation methods. Recursive convolutional neural network not only has the advantages of automatic feature extraction, but also effectively captures the features before and after the time series. However, the traditional video recommendation process has the problems of low time efficiency and low accuracy. To solve the above problems, this paper proposes a method that is based on convolutional recurrent neural network in the barrage video recommendation. Firstly, according to the number of barrage videos, the method selects the preferable video fragments of users, and adopts the k-means clustering method to extract key frames from video. Secondly, we use a convolution and recursive neural network model (RCNN) to classify the similar video fragments. Finally, the recommendation can be achieved by the similarity between video fragments. Experimental results demonstrate the effectiveness and better performance of the proposed method.