Research On Complex Network Reconstruction Algorithm Based On Random Forest

A complex network of individuals that are independent and interacting is ubiquitous in nature and in our society.The topology of the network is the basis and premise for studying complex networks.How to infer the topology of the network from the network dynamic time series is an attractive problem,and it is also a major challenge in many research fields such as mathematics,biology and engineering.This thesis mainly studies the use of a class of machine learning algorithms to infer the internal structure of the network through the time series data of the network nodes.In particular,this thesis infers the directed topology of the network by studying the driving causality between different nodes.This thesis establishes the basic model of dynamic behavior of complex networks,and thenbuilds an algorithmic framework based on the learning algorithm of random forests,so that the directed causality of the network can be characterized and measured by the impurity of random forests,Finally,the topology reconstruction of the entire complex network is realized,and the reconstruction effect of the network is represented by the area AUC surrounded by the ROC curve and the coordinate axis.In this thesis,the application results of the algorithm framework on the verification model and the gene regulation network dataset are given.The reconstruction effect is compared with the compression-based reconstruction algorithm,and the influence of parameters such as time series length on the performance of the algorithm is discussed.The robustness of the algorithm in different noise environments is given,which verifies that the algorithm framework has wide applicability and high stability.