基于社会网络分析的多属性关联规则挖掘方法

A Method for Mining Multi-attribution Association Rules Based on Social Network Analysis

应用社会网络分析的方法解决多属性关联规则挖掘的问题，这是解决这类问题全新的视角。首先，从啤酒的不同品牌与尿不湿不同颜色的搭配引出了多属性关联规则挖掘的问题，并指出这类问题也包含着广泛的评价和推荐问题；而后，基于社会网络分析的视角，建立了相应的图模型及与之等价的矩阵，通过对图和矩阵的分析，引出了多属性关联规则挖掘的方法；为了进一步使方法有助于程序化表达，将既有的方法通过引入“指标向量”实现了统一表达，这有助于程序递归的实现；最后，给出了本文方法的算法步骤，并将其应用在一个100000评估量规模的数据集上对方法进行实证分析。结果表明：本文通过社会网络分析的视角将抽象的关联规则挖掘变得可视化，这便于矩阵表达的引入，使得到的方法具有算法复杂度低、直观和易于把握的特征，相比于既有的多属性关联规则挖掘算法有优势。

It is a new perspective to mine multi-attribution association rules by using the social network analysis. First, the problem of collocation between the brands of beer and the colors of diaper led to the concept of multi-attribution mining, which included wide range of evaluation and recommendation problems. Second, the method was created through the analysis of the graph and the equivalent matrix established from the social network analysis. Third, for the realization of procedural expression, an ＂index vector＂ was introduced to unify the formula of the above method, which helped recursion in the program. Forth, this paper made an empirical analysis upon the dataset with 100 000 evaluations by using the method and its algorithm. In conclusion, the application of social network analysis visualizes the abstract mining process, facilitates the introduction of matrix expressions and precedes the existed multi-attribution methods in the aspects of algorithm complexity, intuitive feature and ease to master.