Link Prediction For Composite Networks

In recent years,as the rise of complex network technology,link prediction,a hotspot of complex network research,is also widely used in many fields.Link prediction can be divided into local methods considering node attribute information and global methods using network structure information.Local method is simple to calculate,but there are some drawbacks that data is difficult to obtain.The global approach is considered to be comprehensive and intensive,but the computational complexity is very high.Focusing on the semi-local method,disease gene discovery and haze fluctuation pattern are studied based on the link prediction of composite network.The main works involved in this paper are:(1)The existing method neglects the interaction asymmetry between the nodes,which may not be able to distinguish the similarity between nodes with the same weight and length but different weight distribution.So,in this paper,based on Itan's indicator,an asymmetric similarity indicator-IA(Interaction Asymmetry)is proposed.With the experimental verification on Retinal Pigment(RP)variants data set,IA indicator achieved a better prediction effect,which proved the effectiveness of this indicator.(2)In order to study the factors affecting the occurrence of haze and its mechanism of influence,based on the data of air quality in Shandong Province for more than three years,the haze fluctuation law in Shandong province was studied.In this paper,a complex network is proposed for the haze problem,which is based on the composite network.A time series link forecasting model based on composite network is proposed,which provides the basis for the analysis of time-related correlation with weather factors.