Stock Price Prediction And Analysis Based On EEMD And XGBoost Algorithm

China's stock market,as an important channel for real enterprises to raise funds and various types of investors to allocate their own assets,is an indispens-able part of China's financial market.As a financial asset with high return and high risk,stock plays an important role in the development of the financial mar-ket,and has been attached great importance by people from all walks of life in the financial market.However,stock data is a financial time series with high complexity,high data volume and high frequency of change.At the same time,it has the characteristics of high noise,non-linear and non-stationary,which goes against the basic assumptions of some classical theories and restricts the appli-cation scope of the classical theories.Therefore,it is of practical and theoretical value to find a reasonable method to extract the stock market data and build a model that can describe the nonlinear characteristics of the complexity of the stock market,so as to further reveal the internal operation law of the stock mar-ket,better play the due functions of the stock market,and more timely expose the financial risks.Up to now,the method of model analysis has been widely used in the field of financial industry analysis,among which ARMA,arch,GARCH and other classic models are used.All these methods are based on the basic principles of statistics,and the modeling analysis is carried out on the premise that the time series satisfies the hypothesis of stationarity or normal distribution.These analysis methods are only from the time domain,which has low computational efficiency,insufficient prediction accuracy and single research level.Therefore,the classical forecasting model has been difficult to adapt to the current demand for financial time series research.With the establishment of machine learning system and its wide application in empirical analysis,machine learning algorithm has gradually entered people's field of vision,and injected new power into the research of financial time series prediction.Among these algorithms,xgboost,as an improved algorithm of gbdt,has the characteristics of low computational complexity,fast running speed and high accuracy.The application of xgboost model in the field of stock forecasting can not only improve the forecasting accuracy,but also improve the forecasting speed,which is a very effective learning algorithm.In addition,there are many physical signal processing methods in the engineering field,such as set empirical mode decomposition(EEMD),which are also very effective in the financial field.Based on the above background,this paper introduces the basic principle of EEMD decomposition method and xgboost algorithm in detail.By combining the two methods,the eemd-xgboost composite model is constructed,and the daily closing price series of Shenzhen composite index from January 5,2018 to December 29,2019 are predicted and analyzed,so as to obtain a more optimized stock price prediction model.