Stock Price Prediction Based On Recursive PCA-local Constant Estimation Model

With the rapid development of the stock market in China,more and more people begin to invest in stocks.However,due to the large number of stocks and the frequency of trading,investors need to analyze large amounts of data to get more accurate prediction results.Therefore,in order to reduce the cost of the analysis process,researchers are interested in online algorithms for stock price prediction.Online algorithms are a kind of algorithms that can update the estimate in real time with the arrival of a new data.This kind of algorithms have higher computational efficiency and can lower the requirements of hardware.From another point of view,when the online algorithm is used to analyze the stock market with massive data,it can reduce the costs required for the process,and hence it has certain theoretical significance and application prospects.Based on the theory of stochastic approximation,a principal component analysis method which can satisfy the requirement of online computation is introduced,and a recursive local constant estimation method is improved,which not only fulfils the demand of online calculation but also owns similar performance with its offline version.Then some asymptotic properties of the improved recursive local constant estimation algorithm are given.Finally,the feasibility of these two algorithms is verified by numerical simulation.In the empirical analysis part,this paper extracts the daily closing prices of two stocks in the US stock market in a period of five years.With the data,we employ the proposed two online algorithms to analyze and predict them,and compare the prediction results with the traditional offline algorithms.It is concluded that the recursive local constant estimation algorithm can satisfy the demand of online computing,whose prediction accuracy is not lower than that of some traditional offline algorithms on the selected data.Therefore,it can be considered that the online algorithms studied in this paper have certain practical significance and development prospects for analysis of stock market prices.