【作者】 孙国荣；

【导师】 徐沛娟；

【作者基本信息】 吉林大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 近年来,国内移动通信产业获得长足的发展,用户数量和网络规模迅速扩大,网络通信质量愈发受到重视。目前,网络管理系统每天收集海量的数据以供分析人员进行各种数据和业务分析;为了对海量数据进行科学的分析处理,以保证对网络扩容及优化提供准确的决策支持,将数据挖掘技术引入网管系统中已经成为一个新的研究课题。本文通过深入了解话务网管的领域知识以及对数据挖掘技术的学习和研究,初步设计了数据挖掘技术在移动通信话务数据分析中的应用系统,本文的主要工作如下:1.介绍了数据挖掘技术的基本的概念及技术和话务网管领域基本模型及通信知识;2.从数据挖掘技术入手,结合话务网管领域知识,分别介绍了概念描述、分类、关联、时间序列算法的实现及其在话务网管数据分析中的应用研究;3.分析话务网管的业务需求,为数据挖掘设计了一个数据挖掘数据库库的逻辑模型;4.设计了一个话务网管数据挖掘分析系统NMMDS,本文对这个系统的体系结构、数据模型以及各个功能模块都做了详细的介绍。更多还原

【Abstract】 Along with the mobile service thorough people daily life each aspect, thecommunication network is expanding day by day, service number adding,market competition increasing day by day, various operations business needsthe further understanding the run condition of the entire communicationnetwork, and need to obtain from the massive data for the network planningand the optimized policy-making knowledge, to provide the general mobilephone user a more unobstructed communication network. The traffic networkmanagement system is a comprehensive network management system. It hasused some traditional data statistical methods to provide many kinds ofstatistical data and the report for the user. But facing the massive data,applying tradition methods to detect some concealed knowledge is extremelydifficult. In the recent ten years, data mining technology has developed veryrapidly. It has successful applied in many businesses. The actual demand ofin the telephone traffic network management system needs to apply the datamining technology.The data mining technology is a process that finds the latent usefulinformation or knowledge which people beforehand does not know frommassive, not incomplete, noise and fuzzy practical application data.The data mining divides into two phases, the data preparation and datamining. The data preparation phase is a very important phase in the entiredata mining process. According to data characteristic in network managementsystem, we need to filter and convert the raw data. The data mining is theessential process in the entire data mining process. The data mining modelmay divide into two categories according to its function: predictive modeland descriptive model.Applying data mining technology basic concept, design model,excavation process and correlation data mining technology, I preliminarydesign a traffic network management data mining system (NMDMS).Through the study model data mining system structure module and the actualfunctional of the traffic network management system I have designed specificstructure module. Through the analysis traffic network management systemservice and the new demand, I have designed the function module. Accordingto the actual demand situation, the function module divides into sevensub-modules: The comprehensive resource survey analysis subsystem, theoverall performance survey analysis subsystem, the resource early warninganalysis subsystem, the performance trend analysis subsystem, theperformance warning is connected the analysis subsystem, the KPI quotaconnection analysis subsystem, the network Yuan performance finenessanalysis subsystem.The data mining system primary task is to extract the connotativeknowledge from the database primary information, the connotativeknowledge is what that the people beforehand does not know, but also is thelatent useful information and the knowledge. Through some agreementformats data packets, the data is interacted in between system interior eachmodule. Each module receives the data packet and completes thecorresponding treating processes. Here, I Described three layers flow chart:The top layer data flow chart, the 0-th layer data flow chart and the 1st layerdata flow chart. The relations between various modules in the system aredescribed clearly.The data mining data comes from many systems, so it is necessary todesign an independent database (NMDMDB). The NMDMDB data comesfrom many database, mainly comes from the traffic network managementsystem database (NPMDB) and the warning database. According to the need,we need fixed time, to carry on the data according to the certain cycle thesynchronized work, at the same time we must to the data which goes intostorage carry on the preliminary clean work. The storage must provide themost direct data pool in the NMDMDB database for the data mining.Studying the actual telephone traffic service analysis application andexisting data mining algorithm, I analysis the key data mining algorithm andthe application demonstration in the traffic network management data miningsystem. The class model is difference general character as well as thedifferent races thing of between the feedback similar thing characteristicknowledge. Build some kind of sorter, maps the data set in data to thespecific category on. Its main target is with by extracts the data classcharacteristic model, then forecast thing development tendency. In here usmay discovered some before using the classified algorithm not the rule whichdetected, for instance traffic load and call completing rate existence rule. Therelative rule is between the detection mass data mean terms set the interestingconnection or the correlation relation technical method. For instance, in therelaying information table between various quotas possibly has someconcealed connection relations. These concealments relations very arepossibly important to the policy-maker. I here expected obtains between thesetwo quotas through the connection rule data mining each kind of relations.And it carries on the network management decision-making by thisachievement the basis. The time series model basis data along with the timevariation tendency, detected some in the compartment the data relevantprocessing model, the forecast future will be possible for a while the currentvalue the distributed situation. The telephone traffic network managementstatistics performance index has the corresponding change along with thetime change, but its change has some rules and the cycle. Discovering itsexistence the change rule, we can to do predictive on the short-term and thelong-term tendency forecast.Here the data mining system NMDMS is only a research experimentsystem, regarding the system structure and the model. It needs furtherdiscussion. The practical application is the goal to build a data miningsystem , Also needs unceasing improvement in the actual use to make thedata mining system to become the business management personnel theassistant, . A data mining system can be used in an enterprise’s dailyoperation work is a data mining system true success symbol.更多还原