基于分类模型和SVD的协同过滤算法

Collaborative filtering algorithm based on classification model and SVD

协同过滤算法是近年来运用最为普遍的推荐算法,但具有数据稀疏、冷启动的缺点。为解决上述问题,特提出综合奇异值分解(SVD)和分类模型(CM)的协同过滤推荐(CFR)算法(SCC)。首先分别建立基于机器学习的分类模型和基于SVD的协同过滤模型.前者用于获取推荐标签,而后者用于获取拟推荐物品。其次用推荐标签筛选拟推荐物品,并用Top-N的方法得到推荐物品表,实现分类模型与SVD协同过滤模型的融合。邀请多名志愿者体验不同推荐算法系统进行实验对比。实验结果表明,融合算法的准确性最高达61.92%,而满意度相对SVD算法(相对提高20.007%)与分类模型算法(相对提高5.42%)有不同程度的改善,但在数据较少情况下满意度与准确性提升并不明显,所提算法不仅一定程度上解决了冷启动问题,同时具有降低了推荐过程的复杂度。

Collaborative filtering algorithm is the most popular recommendation algorithm used in recent years,but it has the disadvantages of sparse data and cold start.In order to solve the above problems,a collaborative filtering recommendation(CFR)algorithm combining SVD and classification model(CM)is proposed,which is called SCC algorithm.First,a classification model based on machine learning and a collaborative filtering model based on SVD are established respectively.The former is used to obtain recommended tags,and the latter is used to obtain proposed items.Secondly,the proposed tags are used to screen the proposed items,and the top-N method is used to obtain the recommended item list to realize the fusion of the classification model and the SVD collaborative filtering model.Multiple volunteers were invited to experience different recommendation algorithm systems for experimental comparison.The experimental results show that the accuracy of the fusion algorithm is as high as 61.92%;and the satisfaction is improved to different degrees from the SVD algorithm(relatively increased by 20.007%)and the classification model algorithm(relatively increased by 5.42%),but it is satisfied with less data The improvement of degree and accuracy is not obvious.This algorithm not only solves the cold start problem to a certain extent,but also reduces the complexity of the recommendation process.