Data Fusion, Data Mining and Decision Support System: Bank Marketing in the 21st Century

Principal Investigator:
Leung, Yee
Chan, C. F.
Lai, S. K.
Lau, K. N.
Leung, K. S.
Leung, P. L.

Co-investigator(s):
----

Summary:
This project proposes to develop an innovative decision support system (DSS) to take advantage of valuable customer information embedded in bank's databases and frontier computer technology in changing the basis of competition among banks in Hong Kong. This decision support system, which is practically non-existent in Hong Kong, would allow a bank to selectively categorize their customers based on important attributes such as socio-economic background, demographic factors, and life-style preferences. Currently, information of these attributes is available, but it is highly aggregated and may not even be found in bank's databases. This suggests that a bank must concatenate their customer records with the aforementioned attributes, which are normally available in census data. However, cross-tabulating these data manually would be impossible, due to the complexity and significant requirement of time and resources. Thus, we are proposing this DSS to automate the task. Once the system is completed, the bank could use it to classify their customers based on any built-in variables. Imagine if a bank could achieve this pattern of customer segmentation, they could easily develop innovative bank services such as cross-selling, up-selling, product differentiation, and bank marketing. Best of all, the bank could achieve these goals electronically, suggesting tremendous cost-savings, minimal human effort, and significant efficiency performance.

The development of this unique DSS involves several critical steps. First, we need to identify the external databases that suit our needs and fuse them with the bank's customer databases. This task is complicated due to the fact that external databases are incompatible, incomplete, and inaccurate. Hence, we have to spend a lot of time and effort cleaning and enriching these external databases prior ro the performance of data fusion. Next, we need to build several behavioral and mathematical models to determine the decision sequence, pattern, and priority to support customer segmentation processes. As the decision processes involve the evaluation of many fuzzy variables, therefore, applicable artificial intelligent techniques will be incorporated to the development of our decision models. Once these decision models are developed, our next task is to select the appropriate data mining techniques to query the newly integrated databases. Due to the uniqueness of our project, we have to develop new data mining algorithms despite the fact that today's data mining software is already quite complete. Finally, the last step of our project is to integrate all the products derived earlier by building a user-friendly intelligent DSS. This last deliverable will allow the banks to effectively and automatically segment their customers into their groups of preferences.

This project promises to turn the previously hidden treasure within databases to billions of dollars of revenue through knowledge extraction and product/services innovations. Hang Seng Bank has already agreed to participate in the initial stage of the project. The system, the process, the model, and the software we deliver as the end product of the study will generally be applicable to other banks after appropriate modifications. Our team members are experts and practitioners from various disciplines.