Online Social Networks’ Users’ Behavior Modeling And Predicting

The recent rapid development of social network ecosystem has resulted in intensifying social network users online activities and the pursuit of high-quality multimedia content on respective social network platforms.Consequently,online social network(OSN)users behavior has became so much diverse and complex that it ended up presenting a challenge to the traditional network resource management.This thesis aims to analyze OSN users behavior underlying mechanisms and laws from two aspects: network resource management and network service quality through stochastic theory,Markov process,queuing theory,dynamical system,regression analysis,neural network and data mining methods.This thesis takes the impact of user behavior on network resource management and network service quality as the breakthrough point and builds,based on multi-user network resource demand,user random behavior,network resource allocation complexity and diversity and the compound effect of multi-user interactive network services,online social network user behavior model based on dynamical system.Finally,this dissertation predicts online social network user behavior based on regression analysis and neural network method,optimizes network resource allocation and improves network service quality.The main contents of this thesis are listed as follows:(1)This dissertation discusses in depth existing theories and methods from resource allocation,traffic shaping and data flows priority and proposes new approach in conceptualizing different phases of users data forwarding mechanism,and mathematical models for resource and service optimization and prediction.According to current network management and network quality of service,this investigation provides a new and open perspective to solve network recurrent problems of resource management focusing mainly on user behavior for multiple purposes such as traffic prediction and congestion avoidance,user satisfaction optimization and so forth.(2)From social network user behavior and social network architecture perspective,this dissertation models the impact of OSN users dynamics on network service requirements using Markov Chain(MC)and analyzes the underlying network resource management problem and network service prediction at peak hours when users are intensively interacting on respective SN platform.Finally,using field data provided in the China Internet Network Information Center(CINIC)2019 report,we tested and validated the proposed models predictability and revealed OSN users behavior complexity and network resource requirements.(3)Adopting a systemic or holistic approach,the problem of user data packets queuing time and network segment data forwarding reliability is further analyzed based on queuing and stochastic theories and models,as well as Weibull distribution.Then,using data extracted from an overseas network service provider,this thesis verifies the accuracy of the proposed models and points out the impact of users’ data queuing waiting time on network resources consumption and validates network nodes aging process effect on respective network service quality.(4)Considering OSN multi-user composite mechanism from network service demands perspective,this thesis introduces population dynamic framework and OSN users’ network resources consumption model based on respective queuing data to analyze the competition mechanism of online social network data service resource allocation and overall network service management.Finally,we simulated OSN users’ data forwarding process using Matlab and revealed the rules of network resource management under the compound effect of multi-user multi-service stochastic behavior mechanism.(5)From SN ecosystem limited resources viewpoint,based on linear and nonlinear regression analysis,artificial neural network(ANN),multi-layer perceptron(MLP)and network topology,this dissertation models OSN users behavior composite resource requirements for network service optimization and predicting in respect to available resources and network state.Then,using monitored Ethernet V2 subnet and WLAN 802.11 network data,we tested the proposed models predictability and accuracy in optimizing network service and avoiding traffic congestion.The results show that,at peak hours,optimizing network resource allocation is crucial for ensuring the reliability of the system and network service quality in multi-platform multi-user environment.Furthermore,considering user random behavior significantly improved the proposed model accuracy,showing that human factors play an important role in enhancing social network ecosystem sustainability management,improving SN user satisfaction,measuring existing infrastructure efficiency and reliability,as well as understanding SN users’ available services preferences.