摘要
为有效解决目前绝大多数推荐系统存在数据稀疏、个性化低、计算负荷量大等特点,在最近邻(KNN)模型基础上提出一种结合降维的最近邻算法(KNN-DR),利用矩阵分解的方法降低矩阵的稀疏性,过程中融合更多隐式因子并加快运算速度;在皮尔逊相似度基础上添加延伸相似度,进一步克服数据稀疏性问题。该算法有效解决计算复杂度高和推荐效果大众化的缺点。实验结果表明,KNN-DR算法在推荐准确度上取得了更好的效果。
To solve the problems of sparse data,low personalization,large calculation and etc.in most of current recommended systems,based on the K nearest neighbor(KNN)model,a K nearest neighbor algorithm(KNN-DR)was proposed.Matrix decomposition was used to reduce the sparsity of the Matrix,more implicit factors were incorporated and the computation speed was accelerated.The extended similarity was added on the basis of Pearson similarity,which overcame the problem of sparse data.The proposed algorithm effectively solves the shortcomings of high computational complexity and popular recommendation.Experimental results show that the KNN-DR algorithm has better effects on the recommended accuracy.
引文
[1]Ye H.A personalized collaborative filtering recommendation using association rules mining and self-organizing map[J].Journal of Software,2011,6(4):732-739.
[2]XU Mingjie,WEI Chengjian,SHEN Hang.Spark parallelization based on item collaborative filtering algorithm[J].Computer Engineering and Design,2017,38(7):1817-1822(in Chinses).[许明杰,蔚承建,沈航.Spark并行化基于物品协同过滤算法[J].计算机工程与设计,2017,38(7):1817-1822.]
[3]YIN Hang.Study on collaborative filtering in information recommendation system[D].Shenyang:Northeastern University,2012(in Chinese).[尹航.信息推荐系统中的协同过滤技术研究[D].沈阳:东北大学,2012.]
[4]Thorat BP,Goudar MR,Barve S.Survey on collaborative filtering,content-based filtering and hybrid recommendation system[J].International Journal of Computer Applications,2015,110(4):31-36.
[5]Kumar S,Kumar S.An approach for recommender system by combining collaborative filtering with user demographics and items genres[J].International Journal of Computer Applications,2015,7(4):128-130.
[6]Liao H.A new web service model of hybrid personalized recommendation[C]//9th International Conference on Natural Computation.IEEE,2013:868-872.
[7]Lei W,Qing F,Zhou J.Improved personalized recommendation based on causal association rule and collaborative filtering[J].International Journal of Distance Education Technologies,2016,14(3):21-33.
[8]Liu Y.Research on collaborative filtering recommendation algorithm based on SVD matrix decomposition technique and RKNN algorithm[J].Journal of Hunan Institute of Engineering,2015,9(3):34-36.
[9]Huang Y,Gao X,Gu S.UARR:A novel similarity measure for collaborative filtering recommendation[J].Cybernetics&Information Technologies,2013,13(Special Issue):122-130.
[10]Xia JX,Wu F,Xie CS.A novel similarity measure based on weighted bipartite network for collaborative filtering recommendation[J].Applied Mechanics&Materials,2013,266(9):1834-1837.
[11]Said A,Luca EWD,Albayrak S.How social relationships affect user similarities[C]//In Proceedings of the 2010Work-shop on Social Recommender Systems,2010:1-4.
[12]Groh G,Ehmig C.Recommendations in taste-related domains:Collaborative filtering vs.social filtering[C]//In Proceedings of GROL′07.New York:ACM,2007:127-136.
[13]Li YM,Wu CY,Lai CY.A social recommender mechanism for ecommerce:Combining similarity,trust,and relationship[J].Decision Support Systems,2013,55(6):740-752.
[14]Breese JS,Heckean an detal fanpirical analysis of predictive a1gorithms for collaborative F filtering[C]//Proc of the 14th Conference on Uncertainty in Artificial lntelligence,2008:43-52.
[15]WANG Xiaolin,YANG Lin,WANG Dong.New word similarity algorithm research based on HowNet[J].Information Science,2015(2):67-71(in Chinese).[王小林,杨林,王东.基于知网的新词语相似度算法研究[J].情报科学,2015(2):67-71.]
[16]DU Linna,YAN Guanghui,YANG Xiaxia,et al.An improved KNN Chinese text classification algorithm[J].Software Guide,2010,9(2):51-53(in Chinese).[杜琳娜,闫光辉,杨霞霞,等.一种改进的KNN中文文本分类算法[J].软件导刊,2010,9(2):51-53.]