Bank Cost Analysis Based On Data Mining

With the gradually deepening of the financial system reform and the expanding emergence of innovative financial services, the bank industry in China is undergoing a profound transformation never seen before. How to increase the competitive ability in the market and make a better decision rapidly are the problems facing to all the banks. Meanwhile, the recent Informationalization in bank industry accumulates large volume of data in wide domains. Thus, the technology of data mining is accepted popularly to turn these data into the knowledge, which is promising to benefit the bank running and management.The application domains of data mining in bank industry are wide, including client relationship management, risk analysis and profit analysis of financial products. This paper focuses on the bank cost analysis, which provides more direct and sound reasoning for management and decision-making of a bank. It aims to make good use of the latest research fruits from knowledge discovery, find a concrete field in bank running for data mining and attempt to automatic knowledge acquisition for bank decision-making.This paper gives a summary of the research related to the algorithms used in the bank cost analysis, and puts the emphasis on the algorithm of Classification Based on Association (CBA). The main spirit of this algorithm is: only the association rules between condition attributes and decision attributes are mined as classification rules by any association rule mining algorithm, such as FP-growth. The importance of each rule is measured by its support and confidence, and these rules are ranked in the decreasing order of their importance. The final classifier is obtained after the pruning process on the ranked rules.To verify the correctness of our method, this paper compares the performances of different classification algorithms on the data of bank cost analysis. The experimental results show that the CBA outputs the best accuracy on these data. Thus, based on this algorithm a classification model is generated, which predicts the profit rank for a given bank product according to its descriptive attributes and cost information. It also yields some practically valuable rules, which promisingly provide scientific decision-making.