Using Data Mining to Detect Abnormality in Telecommunication

• Main Authors: CHENG FU SHAN, 鄭富山
• Other Authors: WENG SUNG SHUN
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/83036123940813838779

碩士 === 輔仁大學 === 資訊管理學系 === 89 === Keywords：KDD、Data Mining、Fraud、Neural Network、Entropy In the recent years, Data Mining is a top issue in the field of database applications. Data Mining generally means that it utilizes various kinds of methods and techniques to mine data. It analyzes, generalizes, and integrates the past, accumulated and large quantity of historical information to find out the interesting patterns and pick out useful information as the basis of decision-making for the managers. No matter in categories of retailing, Electronic Commerce, finance, telecommunication, web management, medical diagnosis, or others, people have already understood the importance of KDD（Knowledge Discovery in Databases）gradually. Therefore, they begin to dedicate to Data Mining aggressively for creating the real values for the enterprises. However, as stated above, data mining tends to analyze the large quantity of historical data. But in order to apply it in the real life, some information, such as telephone fraud, network interruption, credit fraud and so on, is needed to let the managers know in time for minimizing the possible loss. But these abnormal situations may change frequently. How to apply the Data Mining techniques to accomplish a real time and adaptive system is the main goal of this thesis. This article will be based on the telecommunication data and uses the “Entropy” of Thermodynamics as the main guide for appraising the information capacity in the database. We use the marked normal and abnormal data as the input of the artificial neural network. Through the constant process of training and learning the artificial neural network, we wish to find out each abnormal situation precisely for helping the managers making the best policy to earn the maximum profit for the enterprises.