Research On Topology Construction And Planning System Optimization Of Transport Network

Technological changes in the field of computer and communication have driven the emergence of new services and the rapid growth of user demand.And it have placed higher demands on the carrying capacity of the transport network.How to effectively improve the efficiency of network planning has become one of the priorities and difficulties of network operators.Network planning is a complex subject affected by many factors.The study on it involves many fields and disciplines.Corresponding to the actual project,network planning includes multiple stages such as topology construction,routing arrangement,and device configuration.And each stage needs to undergo repeated evaluation and modification.Among these stages,topology construction is one of the most important tasks of network planning.Efficient and reasonable topology construction has great potential in reducing costs.In recent years,this field has achieved remarkable results in many aspects,providing rich model and theoretical support for the scheme design of this paper.However,the existing research still have deficiencies in mining and utilization of data resources of the current network.In addition,most of the academic results are not combined with actual planning systems.This paper combines the research of topology construction and the actual development experience of transport network planning and design system.With the theory and tools in the fields of network science and machine learning,this paper proposes a topology construction scheme of transport network and verifies the feasibility and effectiveness of it.We apply the scheme to the actual network planning system to fill its gaps in topology construction.The main work of this paper is as follows:1.A transport network topology construction scheme is proposed.This scheme consists of two modules:initial topology construction and link addition.The initial topology construction module abstracts the topology construction process into a classification problem of node pairs.It not only simplifies the process but also makes full use of the current network data resources.And because of this,the initial topology constructed is suitable for the actual engineering needs.The second module proposes a link addition algorithm suitable for the transport network.It solves some problems of the initial topology such as isolated nodes and further improves its performance.2.Verification of the feasibility and effectiveness of the scheme.By comparing seven common classification algorithms,this paper proves that both the random forest algorithm and the Gradient Boosting Decision Tree(GBDT)algorithm are applicable to this scheme.The average accuracy of both algorithms is above 98.3%.At the same time,the scheme does not strictly limit the proportion and splitting methods of the training data set.When the proportion is in the range of 30%to 70%,good experimental results can be guaranteed.And the accuracy fluctuation caused by the data set change is less than 1%.The feature extraction process of the scheme is simple.When only extracting three features(the node distance,Gabriel model and Geometric model),it can also guarantee an accuracy of 98.0%.In addition,the link addition algorithm in the scheme can stably optimize the topology performance.Compared with the actual topology of the current network,the APL is reduced by 17.9%,and the total number of l inks is only 93.4%of the actual topology.It also solves the problem existing in other similar algorithms that the new links are too long.3.Applying the scheme to actual planning system,making up its gaps in topology construction,and verifying the optimization effect of the scheme in terms of service carrying capacity through routing experiments.This paper realizes the connection between the scheme and the planning system by establishing the mapping relationship between topology and network resources.The actual business of the current network was extracted as routing demand,and the routing arrangement experiment was carried out.The experimental results show that compared with the actual topology,the failure rate of routing of the topology built by our scheme is reduced by 1.1%.And the total number of links and average number of route hops are reduced by 8.8%and 4.0%respectively.Our scheme carries more traffic with less links and guarantees less route hops.At the same time,the distribution of service of our scheme is more compact,reducing the use of some long links.By Comparing the weighted topology,it can be found that the scheme maintains a good similarity with the actual topology,which means that our scheme can not only construct a better transport network topology,but also ensure the applicability of the topology in practical engineering.