摘要
针对目标用户近邻集合选择失准的问题,引入可普适性测度非线性关系的关联指标——最大互信息系数(MIC),并以此测度用户间的相似程度。基于某一给定的阈值,为目标用户选择近邻集合,然后以近邻集合作为训练集,构建支持向量机个性化预测模型,对目标用户的感兴趣项目进行打分预测。仿真结果表明,MIC测度较Pearson等测度选择的近邻集合更为准确,并具有对阈值不敏感的优点。
The correlation indicator,Maximum Information Coefficient（MIF）,which can pervasively measure the nonlinear relationship,is introduced to solve the problem of inaccurate selection of the near-neighbor set of the target users.The indicator is employed to measure the similarity between users.First,the near-neighbor set target users is selected based on a given threshold.Then,the personalized SVM prediction model is built with the attained near-neighbor set as the training set to carried out scoring prediction for the interesting items of the target users.Simulation results show that the near-neighbor set selected by the MIF Measuring is more accurate than that selected by the Pearson Measuring,and has the merit of insensitive to the threshold.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2018年第2期558-563,共6页
Journal of Jilin University:Engineering and Technology Edition
基金
国家自然科学基金项目(31772157)
国家星火计划项目(2014GA770015)
关键词
计算机应用
信息推送
相似性测度
模型构建
最大互信息系数
computer application
information push
similarity measure
model-building
maximum mutual information coefficient
