Short-term Price Prediction Of Financial Market Based On Wavelet Analysis And Grey Deep Learning Network Algorithm

The investment of financial derivatives is an enduring topic in the financial field today,and predicting the price of financial derivatives by using deep learning is also a popular research direction.However,the changing factors of market prices are very complex,it is still difficult to construct a reasonable and comprehensive model to forecast the price.Based on the sentiment of investing and multivariate data of market fundamentals,this thesis uses deep learning methods to reasonably predict the price of financial derivatives(stocks,gold,oil).First,the original data is denoised and reconstructed by using the wavelet analysis.After that,we use single neural network models to train and predict the data,and evaluate this model.Then two combined neural network models are designed to train,predict and combine the high and low frequency data decomposed by wavelet analysis to achieve the best prediction of effect.Finally,the single model and the combined model are comprehensively analyzed and evaluated to obtain the best prediction model.The results show that:(1)The sentiment of investment can affect the price trend;(2)The deep belief network model has a inferior effect on the prediction of time series data;(3)The long short-term memory model has advantages in the prediction of time series data;(4)the combined model has better prediction performance and can effectively predict the volatility of price,its performance is better than single model.There are many factors affecting the price of financial derivatives,which means that a lot of information is contained in the price.The model proposed in this thesis can excavate the multi-level information of the data,and has a good effect on the price prediction of financial derivatives.