Abstract/Description

Nowadays, the banking sector is increasingly relying on Automated Teller Machines (ATMs) in order to provide services to its customers. Although thousands of ATMs exist across many banks and different locations, the GUI and content of a typical ATM interface remains, more or less, the same. For instance, any ATM provides typical options for withdrawal, electronic funds transfer, viewing of mini-statements etc. However, such a static interface might not be suitable for all ATM customers, e.g., some users might not prefer to view all the options when they access the ATM, or to view specific withdrawal amounts less than, say, ten thousand. Hence, it becomes important to data mine the ATM transactions in order to extract and understand useful patterns concerning the customers' behaviors. In this work, we aim to address this requirement. This paper is the second one (Part II) in a series of two papers (Part I and Part II). In Part I, we have described the selection and pre-processing of an ATM transaction dataset (from an international bank based in Kuwait). We have also described its conversion into the MXML format, in order to data mine it through the ProM tool. In this paper, we import this MXML file into ProM and apply diverse types of data mining algorithms on it. Our results reveal that customers perform money-withdrawing transaction most frequently. Also, it is possible to design adaptive ATM interfaces which cater for the ATM terminal (location) at which the withdrawal is being made, the time of this withdrawal, the number of customers accessing the terminal at this time, and the range of money withdrawn in this time.

Location

Room C5

Session Theme

Artificial Intelligence – I

Session Type

Other

Session Chair

Dr. Sajjad Haider

Start Date

23-7-2011 3:15 PM

End Date

23-7-2011 3:35 PM

Share

COinS
 
Jul 23rd, 3:15 PM Jul 23rd, 3:35 PM

Artificial Intelligence - I: Adaptive Automated Teller Machines - Part II

Room C5

Nowadays, the banking sector is increasingly relying on Automated Teller Machines (ATMs) in order to provide services to its customers. Although thousands of ATMs exist across many banks and different locations, the GUI and content of a typical ATM interface remains, more or less, the same. For instance, any ATM provides typical options for withdrawal, electronic funds transfer, viewing of mini-statements etc. However, such a static interface might not be suitable for all ATM customers, e.g., some users might not prefer to view all the options when they access the ATM, or to view specific withdrawal amounts less than, say, ten thousand. Hence, it becomes important to data mine the ATM transactions in order to extract and understand useful patterns concerning the customers' behaviors. In this work, we aim to address this requirement. This paper is the second one (Part II) in a series of two papers (Part I and Part II). In Part I, we have described the selection and pre-processing of an ATM transaction dataset (from an international bank based in Kuwait). We have also described its conversion into the MXML format, in order to data mine it through the ProM tool. In this paper, we import this MXML file into ProM and apply diverse types of data mining algorithms on it. Our results reveal that customers perform money-withdrawing transaction most frequently. Also, it is possible to design adaptive ATM interfaces which cater for the ATM terminal (location) at which the withdrawal is being made, the time of this withdrawal, the number of customers accessing the terminal at this time, and the range of money withdrawn in this time.