Machine Learning Application In Financial Time Series Prediction

The prediction of time series plays an important role in guiding people's life.With the rapid development of machine learning technology in recent years,its application on time series analysis and prediction has also been broadly developed.However,its effects still need to be studied in many aspects in practical applications and many problems still need to be solved.In this paper,the algorithms of time series prediction have been propose using machine learning technology.This paper mainly contains the following three parts:(1)Improved random forest algorithm on stock yield prediction research,a large number of input features are processed in traditional random forest algorithm,in which al features are unable to be effectively distinguished.Different features of model may have contrary or redundant effects,which means the complexity of the prediction algorithm ia high and even the precision is not very high.A novel algorithm is proposed in the paper using Random Forest algorithm with Particle Swarm Optimization(PSO)algorithm combined with the Grid Search(Grid)algorithm(pso-grid-rf).PSO was used to choose the optimal features into RF,and then GRID was applied to find the optimal parameters synchronously,which effectively prevents the algorithm from falling into local optimization and obtains better classification performance and trend prediction accuracy.(2)The phase-space reconstruction optimization algorithm is applied for stock price regression prediction and time series prediction.It's very important for investor to analyze and model financial time series.However,the non-linear characteristics of financial time series makes it rather difficult to build the model.In this paper,a novel algorithm is proposed using a sliding window time series prediction system,which USES sliding window element optimization to predict stock prices,and particle swarm optimization algorithm to optimize the parameters of support vector machines.Simulation results show that this algorithm can obtain better performance than traditional one.(3)improved support vector machine(SVM)regression forecasting model,many factors,including the history data,technical indicators,market factors,such as stock trading data,the technical indicators of the stock,and the effective factors,affect the future stock price.when many technical features are input as characteristic,principal component analysis(PCA)is applied to effectively choose principal component of stock price,eliminate redundancy between data,shorten the training time,improve the training sample.It's vital for the SVR algorithm to find appropriate parameters.Grid search algorithm is used to search parameter,which proves to besimple and effective.In this paper,a novel algorithm is proposed with the improved grid search of SVR parameters.Simulation results indicate that the effectiveness of the improved algorithm.