Research On Network Operating Situation Prediction Based On Network Service

Communication Networks have developed into large number and high complexity in their sizes,structures,types and amounts of network services currently,which bring big challenges to the administration of networks in the fast-growing of workload and work difficulty.These challenges bring a series of unavoidable problems,such as worse realtime of administration of networks,imbalance in administration of networks and lower accuracy in overall assessment of networks.Under these circumstances,the technology of network operating situation awareness has appeared as a new tech of the overall assessment of networks and the prediction of the overall assessment of networks.This thesis is mainly about the prediction of the network operating situation awareness based on network services and the technology of network operating situation awareness.This thesis mainly solves the problems of how to construct effective situation prediction model,how to achieve the indexes mapping of layers,and how to structure accurate real-time prediction math model.The primary coverage of this thesis is as follows:1.Construction of the network operating situation prediction model based on network services.This thesis constuctures a structure of algorithm of the network operating situation prediction from the indexes of network layer to the indexes of network service layer,then to the network operating situation,based on hierarchical management method.The model combines the analytic hierarchy process based on constrained optimization,expertise and modified BP neural network well,and constructs an effective real-time network operating situation prediction frame.2.The analytic hierarchy process based on constrained optimization.This thesis proposes an analytic hierarchy process based on constrained optimization,aiming at solving the problems of traditional analytic hierarchy process that the analytic hierarchy process requires a high level of the consistency of judgment matrix and it is difficult to adjust judgment matrix to a high level of consistency.The new algorithm solves the problems above well and raises the tolerance of the analytic hierarchy process to the inconsistency of judgment matrix through changing the elements of judgment matrix and using constrained optimization method.It is an advanced indexes mapping of layers.3.The modified BP neural network.The low speed of training of traditional BP neural network has been a weakness of traditional BP neural network since its appearance.Although there has been some modified algorithm from scholars,the effect of these modified algorithm is not very valuable.This thesis proposes a modified BP neural network by working on the root cause of the low training speed,which is called jaggies in steepest descent method.The new algorithm raises the training speed of BP neural network substantially by improving the direction vector of the function updating.