The Application Of Neural Network In Financial Data

Neural network is an algorithm in data mining. Depending on the system's complexity, it can achieve the purpose of processing information through the adjustment of interconnected between nodes internal relations. This paper mainly uses SOM neural network algorithm to cluster analysis, and forecasts the data using BP neural network algorithm,RBF neural network algorithm and general regression neural network GRNN algorithm.The main studying object in this paper is financial data of the China statistical yearbook, and the financial data includes finance income and expenditure. This article mainly aims at researching the central government fiscal revenue of the 31 Provinces,Autonomous Regions and Municipalities, and using the knowledge of neural network, it established the following two research models:(1) According to the regional finance income,the paper divides all the Provinces,Autonomous Regions and Municipalities to three kinds using neural network clustering analysis, and they are first kind region,second kind region and third kind region. The first regions are developed regions, and the third regions are the developing areas.(2) According to the regional financial expenditure of certain areas during several years, it researches the financial data by neural network prediction method. Then it gets the next year's fiscal expenditure budget of the region, and supplies reference for financial budget.Aiming at the financial revenue clustering model, it mainly adopts the following several steps:(1) Getting main factors influencing the financial revenue after multi-element linear analysis to finance income data; (2) Making the main revenue influence factors and total revenue as neural network's input/output respectively, and making cluster analysis; (3) Analyzing regional finance income according to the clustering results obtained.According to the forecast model of fiscal expenditure, it mainly adopts the following several steps:(1) For the fiscal expenditure data, it analyzed factors influencing the financial expenditure, and put these influence factors as neural network's input node; (2) Forecasting the financial expenditure by BP neural network algorithm; (3) Forecasting the financial expenditure by RBF neural network algorithm; (4) Forecasting financial expenditure by general regression neural network (GRNN) algorithm;(5) Comparing the forecasting results got by the above three neural network algorithms, getting the neural network algorithm's prediction accuracy, and analyzing the financial predictions.According to the financial data on fiscal revenues, it builds multivariate linear regression of income data, and the purpose is to extract the factors playing an important role in the revenue, then it analyzed the influence factors of these influence factors as neural network's input nodes. By clustering method, it accords with practice, so it proved the validity of the SOM clustering analysis method, and it also extracted clustering information related to national economy development from the financial data.Depending on the fiscal expenditure data, it makes prediction of next year'fiscal expenditure through the data sample of the 31 Provinces, Autonomous Regions Municipalities'several years revenue, and it uses BP neural network algorithm,RBF neural network algorithm and GRNN neural network algorithm respectively, and the results proved that using GRNN neural network algorithm can get the best prediction accuracy comparing with the other two methods.