Research On Evolution Analysis Methods Of Online Social Networks

With the development of advanced technology and the changing of people's living habits,online dating platform based on Web2.0,online community,online media and so on have beco me more and more popular.On these platforms,people can not only receive and publish infor mation,but also form their own circle in friends by paying attention to their friends.So the net works on the Internet gather a large number of data,and these data can show intention reaction sentiments of people's livelihood and production.The data mining in this field is very importa nt.In addition,the evolution process happened in the past often indicates the changes of the str ucture and function of network some time in the future.So the study of the evolution process h elps to predict the network evolution trend accurately.In this paper,we study the online social network by the way of data mining and evolution.The analysis method of the online social network is divided into two parts,the traditional algorithm and the new statistical physics method.Online social network will be division of the time or space by the traditional algorithm and we can get the infinite subdivision of network fr agments which consists the final result of the whole network analysis results by fragment anal ysis.Statistical physics new method from the angle of the qualitative analysis of the network e volution reflects the laws of evolution.The form of research includes the macroscopic phenom ena to microcosmic research on the laws of the evolution and properties from micro to macro e volution phenomenon of reasoning.In this paper,we use both of these methods to get the evol ution analysis of online social network in different aspects.The analysis of evolution is mining research in the evolution process,so as to find some potential rules in the evolution process.The evolution mechanism mainly includes co-evolution and self-organization evolution.Co-evolution mainly study in mutual influence between subsystems in evolution process.The subject investigated of self-organization evolution can independently complete the development through the internal mechanism from simple to complex and from low to high self-organization systems.Cooperation and self-organisation theory has its own research method.This paper chooses a traditional algorithm and two kinds of methods of statistical physics to analyse properties of the evolution of social network online.Firstly,the online social network community detection based on transfer learning is studi ed.The nodes tags in online social network are looked as the target data and the related long te xts in social networks are looked as auxiliary data.An algorithm called FSFP(Free Source selection Free Priori probability distribution)is studied,which can transfer knowledge from the lo ng texts to the short.This method extracts the node semantic information as node tag,and it tak es the long text of the network as the source data to make up for the short of the target data se mantics and data sparse.It uses the information online as auxiliary data.Select a learning social media network to label for the node information,build the network undirected graph;From th e graph,extract subgraph that contains all seed characteristic sets.With the help of improved L aplacian Eigenmapsmap,map each node to a low-dimensional space,and get the new feature r epresentation for tags;View the latest label feature representation and target domain data label minimum mutual information of target data as constraint to classify so as to realize online soci al network community based on semantic classification.Experimental results in large data sets show the effectiveness of the algorithm.Secondly,the online social network community discovery based on the improved quantu m genetic algorithm is studied.Based on the same context,the connection of the online social n etwork node can reflect the relationship among the semantic tags.In order to find out this relati onship,an improved quantum genetic algorithm APGA(Apriori-Quantum genetic algorthm)is proposed in this paper.Firstly,the semantic tags of the online social network are extracted by T F-IDF algorithm.Then,the improved Apriori algorithm is applied to mine the association relati onship between semantic tags,and then the relations are transformed to the classification rules.Finally,the optimization classification rules of quantum genetic algorithm is applied.The classi fication rules and the results are eventually obtained and thus realize the semantic community division.The experiments on several data sets show that APGA has a fast running speed and di viding the community with high quality.Thrdly,the master-equation method to analyse the distribution regularity of online social network evolution is studied.In the case of the network satisfying Markov hypothesis,the expression and quantity of network evolution state are determined;build evolution equation based on network state distribution function;the main equations are derived from numerical solution based on the related mathematical theorems;the derivative results show that the probability distribution of network evolution over time;calculate network structure entropy expression with time variable according to the result of derivation.The results of the simulation show that the ordered degree of the network is consistent with the actual situation.Finally,the convergence of the critical points of the online social network evolution base d on the core idea of the mean field is studied.The online social network is extended to the ave rage field,and we build the evolutionary analysis model based on the average field.The contentof the model includes the determination of the order parameter of the network based on the ne twork state distribution function and the driven parameter expressions;the establishment of the order parameter equation of network;determination of the critical value of the parameter driv en according to the network evolution trend;determination of the scaling factor of the distributi on of the order parameter in the vicinity of the critical state.The simulation results show that th e convergence of the scaling factor is consistent with the critical point of the real network.