Research On Network Fault Diagnosis Based On Fuzzy Association Rules

When the network node because of abnormal or fault to form a network alarm,its surrounding network nodes will have a considerable number of network alarm,there is often a certain correlation between these alarm information.How to find the relevance of these alerts,and thus accurately locate the root cause of the network is the core of fault diagnosis,but is also the difficulty.Initially,the expert system has been widely researched and applied in the network fault diagnosis,but it has the insufficiency in the establishment and self-learning of the knowledge base.As the data mining technology is widely used in various research fields,and a lot of research achievements have been made,so the relevant researchers try to explore its application in the network fault diagnosis field,a large number of network fault diagnosis technology based on association rule mining are studied.The expert system and data mining technology combined to solve the problem of knowledge and self-learning,and ultimately get a greater success.Although the introduction of association rules mining in network fault diagnosis has achieved great success,there are still some disadvantages: on the one hand,there is a fuzzy relationship between network alarm and network alarm source,which is not simple determinism mapping relationship,and before the previous approach to ignore this point,the processing method is just hard to divide the corresponding relationship between the network alarm and the root causes of the network alarm,which is bound to have a certain impact on the accuracy of the localization diagnosis of the positioning of the network alarm in the late.On the other hand,because the network has the characteristics of layered,so network alarm in the process of spread is affected by the network layer.Previous methods did not take into account the network alarm and the relationship between the various levels of the network.At the same time,because of the different network equipment suppliers,the alarm generated by network equipment exist certain differences in content and format,to a certain extent,it affect the network alarm correlation analysis.According to the above problem,this paper on the basis of association rule mining technology,combined with fuzzy theory and fuzzy reasoning control technology,studied the network alarm source diagnosis based on fuzzy association rules mining.The main contents of the paper are as follows:1.For the uncertainty between the network alarm information and the non-uniformity of information,it is necessary to establish a unified global network alarm information model,According to the characteristics of the alarm and the relevant rules to extract and quantify the key attributes,to establish network alarm information model.In order to reflect the Alarm is affected by network level at the same time,introduce the Alarm Type attribute.2.For the traditional fuzzy clustering algorithm FCM in the network alarm information fuzzy processing,because the cluster center is generated by random initialization,making the clustering center value is unreasonable,which easily lead to the algorithm into the local optimal and the problem of fuzzy evaluation interval inconsistency.For this reason,through to improve the generation strategy of the initial clustering center matrix,so as to optimize the FCM.Using the improved FCM to fuzzy process the network alarm,and the fuzzy alarm model is formed finally.By introducing the fuzzy membership degree to describe the fuzzy relationship between network alarm,which is different from the traditional Boolean logic representation.3.Because this paper is based on the fuzzy association rules to analysis the rules of network alarm,but the association rule mining algorithm is dealing with data objects that require transactional data,so I need to get in front of the beforehand fuzzy network alarm for transactional processing.This paper through the sliding window mechanism for transactional processing,in order to meet the needs of rule mining analysis,form the fuzzy alarm transaction libraries.4.In the process of fuzzy association rule mining,there will be a phenomenon that the fuzzy support count is plummeted as to the high times.If we still use static minimum support degree,it will make some frequent items were missing,thus losing some strong association rules.Therefore,this paper introduces the idea of dynamic updating of minimum support,and implements DFARM(dynamic minimum support fuzzy association rule mining)algorithm.Finally,this paper combined with the BARM(Boolean association rule mining)algorithm,through the fuzzy and non-fuzzy two alarm transaction database for experimental simulation,performance comparison analysis,highlighting the hard division problem.5.This paper analyzes the important components of the fuzzy reasoning module in detail,and analyzes the forward reasoning driving strategy and the reverse reasoning driving strategy,and explain the blur reasoning results.Finally,through the relevant experiments,a variety of reasoning combination performance testing is carried out.Finally,to obtain the optimal combination of reasoning and fuzzy Hamming matching operator synthesis Trip-I,cooperate to drive forward reasoning strategy.Finally,through the test can accurately identify the root node of the network fault alarm.