Research On Cache Strategy Oriented To Recommendation Diversity

The caching algorithm is an important factor that affects the performance of the video caching server.Due to the development of computing power of edge nodes and the increasingly abundant user information,caching based on recommendation ha become a new research trend.Compared with traditional caching methods,the recommendation based caching algorithm can make recommendations to the user from the cached content of the caching server when the cache misses the user request.Then the user will select a content from the recommendation results with a certain probability to watch,so that the caching server can obtain an additional chance of hitting.Whether the recommendation results are appropriate or not affects the probability that the user selects content from the results to watch,which in turn affects the revenue of cache system.Existing recommendation based caching algorithms conventionally consider the correlation between user requests and results when making recommendations.Compared with the video source server,there are only a small amount of contents stored in the cache server for recommendation,so it's difficult to obtain sufficient relevant recommendation results,resulting in a poor user experience.In addition,the relevance-based recommendation methods will result in a higher degree of redundancy in the recommendation results,which also worse the user experience.In order to improve the user experience when recommend from the cache,in this paper we proposed a diverse recommendation based caching algorithm.In the recommendation process,not only the accuracy of the recommendation results,but also the diversity of the results are considered to provide users with varied contents.Thus alleviating user's recommendation experience loss caused by the low relevance of the results and content duplication to obtain higher cache revenue.Compared with the existing methods,we proposed a cache optimization problem with diverse recommendation,and take the recommendation experience score based on the cache hit rate and the diversity of recommendation results as the optimization goal.The cached contents are generated by maximizing the score to optimize user experience.Specifically,the main works of this paper are as follows:(1)A video category based diverse recommendation method is proposed to improve user experience when recommend from the cache.In the case that the cached content in the cache server is predetermined,we described a model of diverse recommendation from the cached content of caching server in response to user requests to optimize user's recommendation experience.In order to solve this model,the algorithm we proposed considers three factors of content relevance,category diversity and user preference as indicators that afect the recommendation results.The importance of each factor is adjusted by way of function mapping to achieve different cahing goals.The simulation results show that our proposal greatly increases the diversity of recommendations and improves user experience at the cost of a certain cache hit rate drop.(2)A cache generation strategy based on diverse recommendation is formulated.We came up with a method of mixed execution of relevance based recommendation and diversity based recommendation.On this basis,a cache generation model is proposed to maximize the overall user experience in the case of different amount user requests.The computing power of the caching server is allocated according to the amount of user requests to solve this model,so that the server can response user requests in different recommendation methods.The simulation results show that our proposal can greatly improve the user experience score,even if the mixed execution of recommendation methods is applied to the baseline model,the user experience score can also be greatly improved.