Decision Tree Learning Algorithm In The Financial Self-service Equipment Monitoring System

Self-service equipments, such as Automatic Teller Machine (ATM), have been used for many years. They've extended the service-time and service-zone, enhanced the efficiency and reduced the running cost of the banks. As the banks depend more deeply on self-service equipments, higher requests are offered. So it is important to have a reliable and efficient Self-Service Equipment Monitoring System (SSEMS) for many of the banks.Decision tree learning, which is the most frequently adopted method in the machine learning, can directly manifest the characteristic of the data besides to be easily understood. Moreover, the decision tree learning, owning the ability of classification and prediction, can draw the decision rule conveniently. After introduced the decision tree learning method to SSEMS, the running data of the primary modules of the self-service equipments can be analyzed, and the malfunctions can be predicted and discovered in time.Firstly, this paper studies the principium of decision tree, the working process, the criterion of evaluation and the problems which can be dealt with. It synoptically introduces the ID3 algorithm and the algorithms based on it, such as C4.5, CART, SLIQ, SPRINT and PUBLIC. Then it uses the priori-knowledge from self-service equipment's running data to optimize the algorithm and puts forward an improved algorithm, which is called Priori-Knowledge Optimized ID3 (PKO-ID3) algorithm. The improved algorithm enhances the proportion of the important attributes and decreases the non-important ones. PKO-ID3 algorithm transforms the weighted sum to the weighted sum optimized by the priori-knowledge. In the process of building decision tree, it ensures the data which was fewer but important not to be submerged.Finally, PKO-ID3 algorithm was used in the Self-Service Equipment Monitoring System of Bank of Communications, Zhengzhou Branch. Both the theoretical analysis and the experimental comparison show that PKO-ID3 algorithm has better improved performance than ID3 algorithm and expresses a good result for prediction.