情报学报
情報學報
정보학보
2014年
8期
844-857
,共14页
社交网络%社交关系强度%机器学习%数据挖掘
社交網絡%社交關繫彊度%機器學習%數據挖掘
사교망락%사교관계강도%궤기학습%수거알굴
social networks%strength of social relationships%machine learning%data mining
本文利用监督学习的方法从社交网络的用户数据中分两个阶段挖掘最佳的社交关系强度分类模型，并进一步探讨不同用户数据对于社交关系强度的区分能力。研究发现，基于贝叶斯网络算法的分类模型在区分强社交关系的过程中被证明最有效，而基于 Logistic 回归算法的分类模型则在区分出临时社交关系的过程中表现最佳。研究还通过属性分析发现互动性因素总体上对社交关系强度的区分能力最为突出，相似性因素中的共同好友数也有很好的区分能力，但时间性因素对于社交关系强度的区分能力没有被发掘出来。
本文利用鑑督學習的方法從社交網絡的用戶數據中分兩箇階段挖掘最佳的社交關繫彊度分類模型，併進一步探討不同用戶數據對于社交關繫彊度的區分能力。研究髮現，基于貝葉斯網絡算法的分類模型在區分彊社交關繫的過程中被證明最有效，而基于 Logistic 迴歸算法的分類模型則在區分齣臨時社交關繫的過程中錶現最佳。研究還通過屬性分析髮現互動性因素總體上對社交關繫彊度的區分能力最為突齣，相似性因素中的共同好友數也有很好的區分能力，但時間性因素對于社交關繫彊度的區分能力沒有被髮掘齣來。
본문이용감독학습적방법종사교망락적용호수거중분량개계단알굴최가적사교관계강도분류모형，병진일보탐토불동용호수거대우사교관계강도적구분능력。연구발현，기우패협사망락산법적분류모형재구분강사교관계적과정중피증명최유효，이기우 Logistic 회귀산법적분류모형칙재구분출림시사교관계적과정중표현최가。연구환통과속성분석발현호동성인소총체상대사교관계강도적구분능력최위돌출，상사성인소중적공동호우수야유흔호적구분능력，단시간성인소대우사교관계강도적구분능력몰유피발굴출래。
This paper explores the best classification models of the strength of social relationships in two steps with the supervised Machine Learning methods and the user data in Social Networks, and investigates the distinguish ability of different types of user data to the strength of social relationships.The findings indicate that the classification model based on BayesNet algorithm is proved to be most effective when distinguishing the strong social relationships,and the classification model based on Logistic Regression algorithm has the best performance when distinguishing the temporary social relationships.Moreover,with the help of attribute analysis,the study finds that the interactivity factors have the most prominent distinguish ability to the strength of social relationships,and the common friends number in similarity factors also has a good distinguish ability.But,the distinguish ability of the time factors has not been excavated.
