Research On Association Rule Mining Algorithm For Congestion Diagnosis Of Private Network

As the scale of the network becomes larger and larger,the topology of the network becomes more and more complicated as the number of users increases.Especially in the management process of the private LAN,the system administrator needs to analyze and deal with the problem on the basis of grasping the network operation data and status to ensure the quality of the information service of the private network.Network congestion is the most common of many network failures.In the process of diagnosing network congestion in the private network,traditional solutions rely on time-consuming manual methods.These have led to the introduction of new technologies for network congestion diagnosis.This thesis studies the problem of private network congestion diagnosis.It collects network data information and mines association rules to discover hidden association relationships between data to provide support for network congestion diagnosis.Research on large-scale private network data acquisition methods and data processing.Aiming at the characteristics of the complex structure of the private network,the large overall network size,high density of various network nodes,rapid changes in data transmission,and complex network traffic behavior,a hybrid data collection method was designed.The network data is collected by parallel deployment and mixed collection of data collection nodes.Perform data processing on the collected network congestion data,and deal with the problems of data duplication,unstructured,noise,etc.that appear in the original data.By cleaning,integrating,filtering,and transforming the data,the data can meet the requirements of mining association rules.An algorithm model for mining network association rules based on Apriori and FP-Growth is established.An improved algorithm is proposed and an algorithm model is established.Analyze network congestion,define association rule mining algorithm for network congestion,and establish related models.Aiming at the problem that "pruning" in Apriori-based network congestion mining algorithm requires frequent scanning of frequent itemsets,"pruning" is optimized by counting the count of frequent itemsets in the candidate set;it is determined after "pruning" Frequent itemsets need to scan the network congestion database for several times.An improved method for judging the support threshold is proposed to reduce the number of scans,and an improved algorithm model is constructed.Design experiments and verify the effectiveness of Apriori and FP-Growth-based network congestion data association rule mining algorithm and improved network congestion mining algorithm.By setting different support thresholds and control variable methods for handling different network congestion data volumes,the algorithm execution time is compared,and the algorithm execution efficiency and effectiveness are compared.The experimental results show that the improved network congestion association rule mining algorithm reduces the mining time by 46.15%-89.42% compared to the Apriori-based algorithm when processing the same amount of data with different supports,and reduces the time by 10.64%-48.71%compared to the FP-Growth-based algorithm.Under the condition of processing different amounts of data with the same degree of support,the mining time of the improved algorithm is 70.00%-80.55% less than the algorithm based on Apriori,and 25.00%-44.45% shorter than the algorithm based on FP-Growth.Based on the experimental results,the improved network congestion mining algorithm effectively improves the algorithm efficiency.