|Title:||Detecting Fraudulent Bank Cheque Customers Using Data Mining Technology|
|Other Titles:||The Case of Commercial Bank of Ethiopia|
|???metadata.dc.contributor.*???:||Dr. Million Meshesa|
|Publisher:||Addis Ababa University|
|Abstract:||The ultimate goal of any country wide organization like that of Commercial Bank of Ethiopia is to make maximum effort on delivering effective and efficient services to its’ customer. To attain this objective, organization of such type should decide on the best approaches to adopt. However, this requires detailed and accurate information on the existing organizational systems. Unfortunately, such information might belong at different organization and the data may also need expert effort for using of it. So, for this type of concern data mining make use of techniques and tools to bring out the needed knowledge. Most of the Banking sector have customers of their own. And, those customers are a user of the services given by the banks. Banks have procedures for every service they give in their branches. Although most customers obey the rules and served on the proper way, it is obvious that some other may move on the illegal way intentionally. This kind of customers will bring failure for the bank and a loss for other loyal customer of the bank. One of the service given by bank is cheque, which is exposed for this threat. This research, therefore, tried to figure out a way to handle this problem using DM technique to build a model which predict cheque fraudulent at the time of new customer registration for cheque users. This research also strictly follow hybrid data mining process model to build predictive model that performs better having an aim of classifying fraudulent bank cheque by studying their patterns. The dataset for this study was taken from Commercial Bank of Ethiopia (CBE), a dataset of 6856 rows of customer data and 1154 rows of malicious customer data National Bank of Ethiopia (NBE) is considered for the mining task. The heart of this research is bases on experiment conducted using different classification algorithms with diverse parameters. Basically, three classifications techniques are used such as J48, PART and SMO. For each algorithm 10 fold cross validation and percentage split was tried. A model with the best performance was registered using PART algorithm having accuracy of 86.4% with the default parameter. The model drives interesting knowledge such as, although, customers registered in south Addis district who kept low balance in their account has found to be loyal for cheque transaction. Similarly a customer who is a male in west Addis district whose marital status is not mentioned (had missing value) at the time of registration is found to be cheque fraudulent. The research can go for a better accuracy if a standard data has been given, and the research can move one step if the rules generated mapped to user interface for easy access of the model at hand.|
|Description:||A Thesis Submitted to School of Graduate Studies of Addis Ababa University in Partial Fulfillment of the Requirements for the Degree of Master of Science in Information Science|
|Appears in Collections:||Thesis - Information Science|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.