基于长短期记忆的稀疏数据过滤推荐算法

Sparse Data Filtering Recommendation Algorithm Based on Long and Short-Term Memory

采用目前算法对稀疏数据进行过滤推荐时,没有综合考虑用户的整体评分特征和不同项目的单独评分对数据补全的影响,导致MAE值和RMSE值大、F1值小。提出基于长短期记忆的稀疏数据过滤推荐算法,首先通过相关因子对相似性进行计算,利用云模型将稀疏数据缺失项进行补全,然后采用补全后的数据构建长短期记忆网络,通过长短期记忆网络得到简单优化函数并对其求解,最后建立稀疏数据过滤推荐算法模型,完成基于长短期记忆的稀疏数据过滤推荐。实验结果表明,所提方法的MAE值和RMSE值更小、F1值更大。

Currently,some algorithms ignore comprehensively considering the influence of users’overall rating characteristics and the individual scores of different items on data completion during the sparse data filtering recommendation,resulting in large MAE and RMSE values,and small F1 values.In this paper,an algorithm of sparse data filtering recommendation based on short-term and long-term memory was proposed.Firstly,the similarity was calculated by correlation factors,and then the missing items in sparse data were complemented by the cloud model.After that,the completed data was used to build a short-term and long-term memory network.On this basis,a simple optimization function was obtained and solved.Finally,a model of sparse data filtering recommendation algorithm was constructed.Thus,the sparse data filtering recommendation based on short-term and long-term memory was completed.Experimental results show that the proposed method can get smaller MAE values and RMSE values,and larger F1 values.