异质网络组合元路径节点重要性分析方法

Heterogeneous Network Combined Meta-path Node Importance Analysis Method

异质信息网络中节点重要性排名适用于学术评审评议、搜索引擎优化、推荐系统构建等领域,可以帮助人们更好的理解异质信息网络的节点特征.由于异质信息网络的节点重要性分析依赖语义信息,使用单一的语义信息进行节点分析是非常受限的.针对上述问题提出了一种基于组合元路径的异质信息网络节点重要性排名方法,通过对元路径进行组合的方式更大范围的捕捉异质信息网络中的语义信息,使排名更精准;使用指数加权平均数法确定组合参数的寻优步长并更新组合参数,循环迭代的计算节点的重要性排名直至排名稳定,使排名更可信.通过对AMiner数据集进行实验分析,验证了所提方法在准确度和收敛速度上优于同类方法.

The node importance ranking in the heterogeneous information network is applicable to the fields of academic reviewand evaluation,search engine optimization,and recommendation system construction,which can help people better understand the node characteristics of heterogeneous information networks.Since node importance analysis of heterogeneous information networks relies on semantic information,node analysis using a single semantic message is very limited.Aiming at the above problems,a method for ranking the importance of heterogeneous information network nodes based on combined meta-paths is proposed.By combining the meta-paths to capture the semantic information in the heterogeneous information network,the ranking is more accurate.The exponential weighted average method determines the optimal step size of the combined parameters and updates the combined parameters.The importance of the computational nodes of the loop iteration is ranked until the ranking is stable,making the ranking more credible.The experimental analysis of the AMiner dataset shows that the proposed method is superior to the similar method in accuracy and convergence speed.