| At present the network is complex and synthesized, the reliabilityresearch of communication network becomes extremely necessary topic forcommunication network management. So there is not a unified definitionabout reliability of communication network and the reliability researches arebased on different research point. The development of artificial intelligencedevides a new methed for reliability of communication network. Thereliability of LAN/WLAN integration network is studied based on PSO,SVM, neural network and stochastic Petri net, which the purpose is exploringa new appraisal method of network reliability and researching an effective,advanced and intellectualized method to estimate and forecast reliability ofcommunication network.1. A modification to Linearly Decreasing Weight strategy in standardParticle Swarm Optimization is taken by introducing descending index anditerative threshold, and then a modified particle swarm optimizer usingnon-linear inertia weight. The formulated experiment demonstrated that thereare evident superiorities in computational precision, searching speed andsteady convergence.2. A new method which is definiting reliability of communicationnetwork is proposed based on LAN/WLAN integration network, and then animproved PSO is used to select power value of BP neural network.Reliability forecast model of LAN/WLAN integration network is established.3. Support vector regress method is introduced to reliability forecast ofLAN/WLAN integration network, and then the reliability is predictedthrough establishing SVM method. 4. A model to LAN／WLAN integration network system is carried onbased on stochastic Petri net, regarding communication network which notyet exists, a model method had been found for its reliability estimation.Three reliability forecasts models (BP neural network, BP neuralnetwork base on modified PSO and SVM regress model) are formulated inthis paper, and then superiority and inferiority are analyzed. |
PDF Full Text Request