Research On Optimization Of Routing Algorithm Based On Semi-Naive Bayes

In recent years,machine learning technology has been more widely and maturely applied in various computer fields.It has shown great advantages in many fields such as image processing,natural language processing,and personalized recommendation,and is still moving forward with a rapid momentum.The development of new technologies will inevitably bring about innovations in old technologies,which also provides a new solution for researchers in solving traditional problems in related fields.This article aims to improve the performance of traditional routing algorithms by applying machine learning and other related technologies.Least Loaded routing algorithm is the routing algorithm that has been widely used in recent decades,and it is also one of the best routing algorithms in terms of performance.Therefore,it provides a benchmark for the performance of the new routing algorithm.However,the traditional least-load routing algorithm also has some shortcomings,sometimes it may cause excessive waste of network resources on the link.On this basis,this paper uses a supervised semi-naive Bayes classifier,combined with the traditional least-load routing algorithm,considers the dependency of resource usage on each link in the network,and treats the resources usage status on each link as a single attribute,and introduces the independent dependency strategy of the semi-naive Bayes classifier to find the super-parent attribute of each attribute,then calculates the potential blocking probability between each node pair,and finally achieve the minimum link load and the minimum network blocking probability balances between these two goals to find the best route.Specifically,if a service connection is established through a certain route between node pairs,each time a service request reaches an operating network that continuously accepts and releases dynamic service connections,a corresponding network snapshot is obtained,in which to record the number of network resource units used on each link.The semi-naive Bayes classifier learns the information of the historical network snapshot by iteratively,thereby predicting the potential future circuit blocking probability between each node pair.And the candidate routes serving new requests are based on the link load and the entire network is determined by the potential future blocking probability(if the route is indeed used).Finally,the paper demonstrates the performance of the routing algorithm through simulation experiments,and compares it with the traditional least-load routing algorithm and the shortest path routing algorithm.The results show that the supervised semi-naive Bayes classifier-assisted least-load routing algorithm significantly reduces the blocking probability of service connection requests,and its performance is superior to the traditional least-load routing algorithm and shortest path routing algorithm.