Research On Stock Price Trend Based On Machine Learning Algorithm

Nowadays,the quantitative investment technology based on mathematical statistics model has been fully developed and applied,and investors can get great returns based on this method.With the advent of Al era,machine learning and deep learning algorithms have been more widely applied.These cutting-edge algorithms can process massive data in real time,the model can also achieve good fitting effect,and the generalization ability has been greatly improved.In this paper,machine learning and deep learning algorithms are tried to be introduced into stock market prediction.XGBoost and LSTM are respectively used to predict the adjusted closing price of stocks,and the fitting effect is compared.Firstly,this paper describes the concept,application and model optimization of machine learning and deep learning,and then introduces the concept of data source and feature engineering.The empirical objective is to predict the daily adjusted closing price of the Vanguard Total Stock Market ETF(VTI)using data from the previous N days(i.e.,forecast range=1).The data used in this paper is based on VTI's historical price for three years from November 25,2015 to November 23,2018.Secondly,in the financial time series forecasting uncertainty,complexity and time memory,this paper proposes the use of machine learning algorithms to build XGBoost model and the use of the construction of deep learning algorithm of short-and long-term memory neural network model,the stock data to extract the features as the input of the model,at the same time,the above two models and the last value contrast,moving average method,linear regression method,the contrast of selected indicators for the root mean square error(RMSE)and mean absolute percentage error(MAPE).Finally,the importance of feature scaling to the prediction results of the model is illustrated.The mean value of the sequence set is scaled to 0 and the variance is 1,so as to train the model.Then,when the prediction of the validation set is carried out,each feature group of each sample is scaled so that its mean value is 0 and its variance is 1 to compare the prediction effect of these five algorithms,indicating that XGBoost has the best feature selection and prediction results.