链路预测方法与网络结构的相关性

摘要

作为数据挖掘的一个分支,链路预测经过多年研究,已经提出了多种链路预测方法.基于网络结构的链路预测方法因为在相似结构的网络中具有普适性,在近年来得到了广泛关注.然而,由于每个复杂网络的结构不同,链路预测方法的预测准确度差别较大,同一种链路预测方法不可能在所有网络中都获得理想的预测效果.研究链路预测方法的准确度和网络结构的相关性可以在已知一个复杂网络结构的情况下,选择合适的链路预测方法,也可以为链路预测的改进提供理论依据.通过实验计算出常见链路预测方法在每个网络上的预测准确度,发现链路预测方法的预测准确度在不同结构的网络中是不同的,结合网络的结构对实验结果进行分析,并据此提出了一种简单的链路预测方法选择方案.%As a branch of data mining,link prediction has been studied many years in the computer field,and a variety of link prediction methods are proposed. The link prediction method based on network structure has been widely concerned in recent years,because it is common in the network with similar structure. However,due to the different structure of complex networks,the prediction accuracy of the link prediction method is different,and the same link prediction method can' t get the desired results in all networks. Analyzing the corre-lation of network structure and the accuracy of link prediction method can help to choose a right link prediction method for a given net-work,and provide the theoretical basis for improving link prediction. The prediction accuracy of the link prediction method is calculated by the experiment,and the results show that it exists difference in the different network structure. Analyzing experimental results combined with the network structure,a simple selection of link prediction method is proposed.