Research on Financial Service Data Based on Neural Network

With the gradual increase of transaction data collection frequency in financial market, high-frequency financial data with nonlinear, nonstationary and high noise has attracted the attention of many scholars, and short-term prediction of high-frequency financial data has become a hot research point in recent years. Research based on financial service data can help financial institutions achieve personalized services. By analyzing their customers’ trading, spending and investment habits, financial institutions can customize products and services to provide solutions that are closer to their customers’ needs, thereby increasing customer satisfaction and loyalty. Financial services data research can help reveal the financial needs and behaviors of different populations and help design more inclusive financial products and services. Through universal access to digital financial services, more people can be integrated into the financial system and promote inclusive economic growth. However, the traditional time series model relies on linear regression to explain the relationship between variables, so as to achieve the purpose of forecasting financial series, and cannot well mine the characteristics of high-frequency financial data. The unsupervised learning process of neural networks can be better applied in the research of high-frequency financial data with nonlinear characteristics. With the innovation of computer technology, the structure of neural networks is constantly deepened, and deep neural networks (DNN) with deeper network structure are widely used. In this paper, the neural network is combined with empirical mode decomposition (EMD), and the IMF is grouped and reconstructed according to the fluctuation frequency, and the reconstructed data is trained and predicted. Finally, experiments are carried out to verify the results, and the results show that the network prediction model improved by ensemble empirical mode decomposition is more accurate in predicting high-frequency financial data.