The Research Of ATM Cash Prediction Based On Improved Genetic Algorithm And Neural Network

With the rapid development of the financial electronic business, bank field also has changed a lot. automatic teller machines(ATMs) is the inevitable outcome of the reform of bank. it has a natural advantage on improving the bank service ability, and make up the shortage of bank outlets. However, cash management has become a thorny problem for each bank, that is how to determine the best number of cash in the ATM to meet the uncertain needs of our customers, and reduce the operating costs of ATM machine. Therefore, more and more scholars and experts have paid more attention to cash forecasting on ATM.For cash forecast modeling problem, this article is based on a large number of domestic and international papers, and aslo analyzes the research background and significance of cash forecasting on ATM. what's more, This paper also introduces the process of the cash forecast system for data mining,and the main work in the process——setting up cash forecast model.Finally, this paper demand using BP neural network to build forecasting model by the comparing with Support vector machine(SVM).Then, this paper expounds the content of BP neural network, and mainly put forward two improved methods towards the defects of BP neural network. on the one hand, in order to achieve optimal prediction model structure, it is a good choice to depend on some theoretical method and constantly ajust the number of hidden layer nodes and input layer nodes by training; on the other hand, on the basis of the optimal prediction model structure, the parameters of BP neural network is optimized by improved genetic algorithm which is improved from three aspects, namely using the Halton sequence to initialize the population, improving the fitness function and using adaptive mutation probability. The optimized prediction model has a better predicted precision.At the same time, this article try to set up an effective forecasting model on realities of situation. considering the bank of actual demand, this paper collect lots of real data of an ATM of the bank branch in Jiangsu in order to build a cash forecast model. the programming realization is in matlab where three prediction model was established based on BP neural network. From the experiment, it is very easy to learn that the prediction accuracy of model is greatly improved by adjusting network structure and using improved genetic algorithm.Finally, a summary of this paper were done, and the further research on The follow-up work were discussed.