The Research Of Financial Forecasting Models Based On Deep Learning And Elliott Wave Principle

Financial globalization is an irreversible trend of the development of human beings.The high complexity,high return and high risk of financial market make it have important research value.The purpose of studying the financial market is to find out its underlying principles,take precautions against financial risks and achieve effective regulation of the financial market.Elliott wave principle has made the elaboration to the fundamental principle of the financial market.The theory has important guiding significance to the prediction of future trend of the financial market.The research on financial prediction model combined with Elliot wave principle is not mature,and it is mostly limited to the shallow network model based on BP network.There is a lack of research on deep learning modeling.This paper studies and summarizes recent financial prediction models based on traditional statistics methods and machine learning methods,and financial prediction models based on Elliott wave principle.Deep learning is a breakthrough technology of nonlinear intelligent model,which has been successful and outstanding in many fields in recent years.Deep learning aims to extract the high-level feature representation of data,and then find out the inherent law of data.Elliott wave principle classifies and describes the changing law of financial market by wave modes.This paper considers that Elliott waves are high-level feature representations of financial time series.Taking this as the joint point,this paper proposes prediction model of Elliott wave patterns using deep learning.Deep belief network is one of the classical models of deep learning,which can realize the goal of feature extraction and classification task of data.Feature extraction and classification are the keys to predict the financial market using Elliott wave principle.Therefore,this paper creatively proposes the PVD model(PLR_VIP+DBN)based on the deep belief network.The validity of PVD model is proved by experiments.It can predict the future trend of the financial market by extracting and classifying Elliott wave patterns from financial time series.Considering the comprehensiveness of the research,this paper introduces five kinds of reference models to model the Elliott wave pattern recognition of financial time series,and comprehensively compares the performance of various neural networks.The reference model includes three deep learning models,traditional BP network and its improvement network.In addition,the PVD model is contrast with the previous financial prediction model.The results of comparative experiments show that the PVD model based on the deep belief network is superior to other models in stability,convergence speed,accuracy and other aspects,so it has higher prediction performance.The PVD model can not only effectively improve the local optimum caused by using BP algorithm in batch,but also improve the representation ability of the shallow network models.It can also improve the prediction performance of traditional BP network model and multiclassifier for Elliot wave pattern recognition proposed by Kotyrba et al.It also improves the accuracy of financial forecasting models without Elliott wave principle,represented by stock returns and high-frequency trading strategies.