Abstract: With the increasing scale of the network,a large random distribution between routing nodes occurs.It is difficult to establish the accurate localization model,so the traditional methods based on feature matching is difficult to describe the random characteristics in fault location of nodes,resulting in inaccuracy of fault location.This paper discussed the overview and characteristics of stochastic networks,and degree distribution and characteristic spectrum of stochastic networks.Through defining the diameter and average distance for stochastic network,calculated the clustering coefficient of stochastic network,extracted effective parameters of stochastic networks fault feature,constructed a feature module of fault data.Using the fault feature node of random walk network fault detection algorithm as a random seed registration point and a node failure detection method for the initial,the methods prove the detection efficiency using the combination of random variation of fault characteristics and registration detection.Simulation results show that the improved method can improve the accuracy of fault localization.

Key words: Random walk,Search algorithm,Node,Distribution,Agglomeration