一种改进的协同过滤算法研究——以电影推荐系统为例

Research on An Improved Collaborative Filtering Algorithm——Taking the Movie Recommendation System as An Example

互联网上信息资源的爆炸式增长,给用户带来了信息过载问题,不明确的用户需求更是对搜索引擎提出了更大的挑战。个性化推荐系统实现了用户和信息资源的紧密连接。目前,协同过滤算法是个性化推荐系统中使用最广泛的算法。然而随着用户数量和信息资源的不断骤增,数据不可靠、稀疏性以及及时性等问题严重影响着推荐系统的推荐质量。因此,基于现有协同过滤技术在时间和空间的维度上对传统算法从三方面进行了改进:在空间上构造情感得分矩阵并结合评分矩阵以缓解用户主观评分不可靠问题;在时间上引入时间权重因子模拟用户的兴趣迁移以缓解数据稀疏性和及时性问题;结合用户兴趣分布相似度和评分相似度来进一步保证推荐质量。同时,以电影推荐系统为例验证改进协同过滤算法的推荐质量,结果表明,相较于传统算法,改进的算法其推荐效果更优。

Massive information resources on the Internet has brought information overload problems to users,and unclear demands of users present greater challenges to search engines. While the personalized recommendation system enables a tight connection between users and information resources. Currently,collaborative filtering algorithms are the most widely used in personalized recommendation systems. However,with the continuous increase in the number of users and information resources, problems such as unreliable data,sparsity and timeliness seriously affect the quality of the recommendation system. Based on the existing collaborative filtering technology,this paper improves the traditional algorithm from three aspects in the dimension of time and space. Firstly,we construct the emotional score matrix and collaborate with scoring matrix to alleviate the problem of unreliablity. Then we introduce time weight factor to simulate user interest migration to deal with data sparsity and timeliness. Furthermore,we combine user interest distribution and score similarity linearly to ensure recommendation quality. This paper takes the film recommendation system as an example to verify the recommendation quality of the improved collaborative filtering algorithm proposed in this paper. The results show that the improved algorithm has better recommendation effect than the traditional algorithms.