基于矩阵填充的混合协同过滤推荐算法研究

Research on Hybrid Collaborative Filtering Recommendation Algorithm Based on Matrix Filling

针对传统协同过滤算法中由于评分数据稀疏性而导致推荐效率低下问题,提出一种混合协同过滤推荐算法.算法首先使用Slope One算法计算出预测评分,以填充原始评分矩阵中未得到评分的项目,降低了评分矩阵的稀疏度,保证了填充值的多样性.然后使用SVD技术对填充的评分矩阵奇异值分解,采用随机梯度降低误差的方法,对结果进行分析,寻找最佳效果参数.最后以五折交叉实验的方式在在MovieLens数据集中对本文所提出的算法与传统的协同过滤算法进行了验证,实验显示算法能够有效地缓解数据稀疏性问题,并且具有更佳的推荐效果.

For the problem of low recommendation efficiency caused by the sparsity of rating data in traditional collaborative filtering algorithm,this paper proposes a hybrid collaborative filtering recommendation algorithm.Firstly,slope one algorithm is used to calculate the prediction score to fill in the original score matrix,which reduces the sparsity of the score matrix and ensures the diversity of the filling value.Then,SVD technology is used to decompose the singular value of the filled scoring matrix,and the random gradient method is used to reduce the error.The results are analyzed to find the best effect parameters.Finally,the proposed algorithm and the traditional collaborative filtering algorithm are verified in MovieLens dataset by 50% crossover experiment.The experiment shows that the algorithm can effectively alleviate the problem of data sparsity,and has better recommendation effect.