Financial Market Prediction and Simulation Based on the FEPA Model

Abstract

Since the birth of the financial market, the industry and academia want to find a method to accurately predict the future trend of the financial market. The ultimate goal of this paper is to build a mathematical model that can effectively predict the short-term trend of the financial time series. This paper presents a new combined forecasting model: its name is Financial Time Series-Empirical Mode Decomposition-Principal Component Analysis-Artificial Neural Network (FEPA) model. This model is mainly composed of three components, which are based on financial time series special empirical mode decomposition (FTA-EMD), principal component analysis (PCA), and artificial neural network. This model is mainly used to model and predict the complex financial time series. At the same time, the model also predicts the stock market index and exchange rate and studies the hot fields of the financial market. The results show that the empirical mode decomposition back propagation neural network (EMD-BPNN) model has better prediction effect than the autoregressive comprehensive moving average model (ARIMA), which is mainly reflected in the accuracy of prediction. This shows that the prediction method of decomposing and recombining nonlinear and nonstationary financial time series can effectively improve the prediction accuracy. When predicting the closing price of Australian stock index, the hit rate (DS) of the FEPA model decomposition method is 72.22%, 10.86% higher than the EMD-BPNN model and 3.23% higher than the EMD-LPP-BPNN model. When the FEPA model predicts the Australian stock index, the hit rate is improved to a certain extent, and the effect is better than other models.

Ardp: Simplified Machine Learning Predictor for Missing Unidimensional Academic Results Dataset

“…However, because these data are about humans, whose behavior is largely unpredictable due to other difficult-to-predict factors, using them as a barometer for prediction yields very few or no correlations. These were highlighted by McCalla (2004), Castro et al (2007), Koedinger et al (2008), Baepler & Murdoch (2010) and Zhou (2021). For example, a large volume of data is discarded because users don't really know how to get the best meaning out of it.…”