Research On Sparse Linear Method Based On Cascade Filtering For Living Streaming Channel Recommendation

An algorithm of living streaming channel recommendation is to meet the needs of personalized viewing of viewers on living streaming platform and the needs of interests of living streaming platform or anchor.After analyzing the current researches on living streaming channel recommendation algorithm,it is found that these algorithms were limited to the viewing behavior when analyze viewer preference,ignored the impact of inevitable noise data and could not keep fast training speed and good recommendation performance in the face of huge and sparse data.Therefore,it is theoretical and practical significance to design a living streaming channel recommendation algorithm with higher accuracy and practicality.In order to solve the limitation of the existing living streaming channel recommendation algorithm on viewer preference analysis,data onto kinds of behaviors of living streaming viewers was analyzed,and it was considered that the viewing behavior,the barrage sending behavior,and the gift giving behavior are important to the expression of viewer preference.Based on this,the viewer behavior matrix is constructed.In order to solve the problem of the existing algorithm on ignoring the noise data,after analyzing the living streaming channel classification label data,it was considered that the rough classification label data could be used to assist in filtering the noise data.Based on this,the channel label matrix is constructed.Based on the analysis of viewer behavior data characteristics,it is assumed that the more they like a channel,the more energy will be spent on the channel,and the influence of different behaviors on their preference expression is different.Based on this,a living streaming channel preference evaluation model is constructed.In order to achieve high recommendation accuracy,noise data filtering and other purposes,combined with the advantages of sparse linear method(SLIM)in the face of huge and sparse data,and the characteristic of cascading hybrid based on classification labels,sparse linear method based on cascade filtering for living streaming channel recommendation was designed.Sparse linear method based on cascade filtering for living streaming channel recommendation is compared with the latest living streaming channel recommendationalgorithms.Indicators such as hit rate,average reciprocal hit-rank,and coverage are designed.The results show that its comprehensive performance is better than other algorithms which means the living streaming channel recommendation algorithm proposed achieved higher recommendation accuracy and covered more living streaming channels in results.Therefore,sparse linear method based on cascade filtering for living streaming channel recommendation improved the recommendation accuracy and achieved the practical purpose of increasing the diversity of recommendations.