基于时间权重因子的双向个性化推荐仿真
Simulation of Bidirectional Personalized Recommendation Algorithm Based on Time Weight Factor
吴翔 1郭飞雁2
作者信息
- 1. 长沙工业学院计算机学院,湖南 长沙 410000
- 2. 湖南科技大学计算机科学与工程学院,湖南 湘潭 411100
- 折叠
摘要
双向个性化推荐需要获取到用户和物品之间的交互行为数据,由于在推荐时,通常只考虑用户的兴趣偏好,忽视了用户兴趣会随时间而变化的情况.为了提高用户满意度,令需求和供给准确匹配转化为网站价值,提出了基于时间权重因子的双向个性化推荐算法.关联用户与邻近用户的关系,分析其商品偏好,计算出用户间的购买相似性,获得相似的推荐用户集合.按照余弦相似度确定用户群的商品偏好权重,计算购买差别,获得用户的偏爱标准化权重.将逻辑斯谛函数引入到皮尔逊相关相似度,判断用户间的相似度,推荐用户偏好的类似商品信息,实现双向个性化推荐.实验结果证明,所提算法能够准确得知用户兴趣的变化,且推荐平均绝对误差小.
Abstract
Bidirectional personalized recommendation requires obtaining interaction behavior data between users and items.However,during recommendation,only user interests and preferences are usually considered,ignoring the fact that user interests may change over time.In order to improve user satisfaction and accurately convert the demand and supply into website value,this paper put forward a bidirectional personalized recommendation algorithm based on time weight factor.Firstly,the relationship between users and adjacent users was associated,and their commodity pref-erences were analyzed at the same time.And then,the purchase similarity between users was calculated to obtain a set of similar recommended users.Secondly,the product preference weights of user groups were determined according to cosine similarity.Meanwhile,the purchase difference was calculated to obtain the standardized weight of user prefer-ence.Thirdly,the logistic function was introduced into the Pearson correlation similarity to determine the similarity between users.Finally,similar commodity information preferred by users was recommended.Thus,the bidirectional personalized recommendation was achieved.Experimental results prove that the proposed algorithm can accurately ob-tain the changes in user interests,and the average absolute error of recommendation is low.
关键词
时间权重因子/双向个性化推荐/偏好权重/排序阶段/余弦相似度Key words
Time weighting factor/Bidirectional personalized recommendation/Preference weight/Sorting stage/Cosine similarity引用本文复制引用
出版年
2024
NETL
NSTL
万方数据