[1]刘金梅,舒远仲,张尚田.基于评分填充和时间的加权 Slope One 算法[J].计算机技术与发展,2021,31(01):35-42.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 007]
　LIU Jin-mei,SHU Yuan-zhong,ZHANG Shang-tian.A Weighted Slope One Algorithm Based on Rating Filling and Time[J].,2021,31(01):35-42.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 007]
点击复制

基于评分填充和时间的加权 Slope One 算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年01期
页码:
35-42
栏目:
大数据分析与挖掘
出版日期:
2021-01-10

文章信息/Info

Title:
A Weighted Slope One Algorithm Based on Rating Filling and Time
文章编号:
1673-629X(2021)01-0035-08
作者:
刘金梅; 舒远仲; 张尚田
南昌航空大学 信息工程学院,江西 南昌 330063
Author(s):
LIU Jin-mei; SHU Yuan-zhong; ZHANG Shang-tian
School of Information Engineering,Nanchang Hangkong University,Nanchang 330063,China
关键词:
矩阵填充; 加权 slope one 算法; 时间因子; FTWSOA; 评分预测
Keywords:
matrix filling; weighted slope one algorithm; time factor; FTWSOA; rating prediction
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 01. 007
摘要:
针对数据稀疏性,常用的评分矩阵填充方法主要是通过平均数、中位数等进行填充,该文提出一种新的评分矩阵填充方法。 利用项目-属性矩阵计算用户对项目属性偏好,由于每个项目都有各自属性,从而可以获得用户对项目的偏好值,以用户平均评分为基准,实现对评分矩阵填充。 基于填充后的评分矩阵,又考虑到用户兴趣爱好随时间会发生改变,因此引入时间因子,提出一种基于评分矩阵填充和时间因子的加权 slope one 算法(FTWSOA)。 通过时间函数修正评分矩阵,优化的评分数据可以更好地体现用户兴趣爱好随时间变化的情况。 在时间加权的评分矩阵下,计算出属性兴趣偏好,在共同评分很少甚至没有时,利用属性兴趣偏好可以较为准确地计算用户相似度。 由于在共同很少或者没有时,原始评分矩阵中用户没有交集,而在属性兴趣矩阵下用户会存在交集,因此,使用参数λ将填充矩阵下的用户相似度和属性兴趣偏好矩阵下的用户相似度相结合得到最终的用户相似度,可以缓解在稀疏数据下相似度计算性能差的问题,最后使用加权 slope one 预测评分时,将时间衰减函数加入到预测公式中来优化预测评分公式。 通过在 MovieLens100k 数据集上的实验表明,相比于其他算法,FTWSOA 算法准确度有所提高。
Abstract:
In view of the sparsity of data,the common filling method of rating matrix is mainly through the average number, median and so on.? ? We propose a new filling method of rating matrix. The item-attribute matrix is used to calculate the user’s preference for item attributes. Because each item has its own attribute,the user’s preference value for item can be obtained. Based on the user’s average rating, the rating matrix can be filled. Based on the filled rating matrix,and considering that user’s interests and hobbies will change with time, a weighted slope one algorithm (FTWSOA) based on the filled rating matrix and time factor is proposed. By modifying the rating matrix with time function,the optimized rating data can better reflect the changes of user’s interests and hobbies with time. In the time weigh-ted rating matrix,we can calculate the attribute interest preference. When there is little or no common rating,we can use the attribute interest preference to calculate the user Similarity more accura-tely. Because there is no intersection in the original rating matrix when there is little or no common,but there will be intersection in the attribute interest matrix. Therefore,the parameter λ is used to combine the user Similarity under the filling matrix and the user Similarity under the attribute interest preference matrix to get the final user Similarity,which can alleviate the poor performance of Similarity calcula-tion in sparse data. When the weighted slope one is used to predict the score,the time decay function is added to the prediction formula to optimize the prediction score formula. Experiments on Movielens100k dataset show that the accuracy of FTWSOA is improved compared with other algorithms.

相似文献/References:

[1]刘金梅,舒远仲,张尚田,等.融合巴氏系数的加权 Slope One 算法[J].计算机技术与发展,2020,30(11):74.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 014]
　LIU Jin-mei,SHU Yuan-zhong,ZHANG Shang-tian,et al.A Weighted Slope One Algorithm Based on Bhattacharyya Coefficient[J].,2020,30(01):74.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 014]

更新日期/Last Update: 2020-01-10