Research On The Application And Prediction Of Financial Time Series Method In Advance-decline Series Of Stock Market

With the rapid development of artificial intelligence and big data technology,the application of machine learning algorithms and deep learning algorithms in data mining in the financial field has gradually emerged.Although the traditional statistical methods in data mining methods perform well when processing conventional data,but there are certain limitations for processing massive amounts of non-positive,non-stationary,nonlinear,and high signal-to-noise ratio financial time series data.Especially the regularity contained in the financial time series is very time-sensitive and the frequency of iteration is high.Traditional models built under harsh assumptions are stretched to deal with massive and complex financial sequences,and the use of machines in data mining algorithms after acquiring new data,learning algorithms and deep learning algorithms can quickly mine valuable information and have strong applicability to the environment.Therefore,it is of great practical significance to explore combining machine learning algorithms and deep learning algorithms in financial time series research.This article uses traditional time series methods,machine learning,and deep learning algorithms to study the grading sequence of the overall increase in China's stock market,digs out the operation rules of the stock market from the sequence,and develops the advancedecline entropy(ADE)index.This has important theoretical and practical significance for quantifying the uncertainty of stock market fluctuations,grasping stock market sentiment,and guiding investment decisions.Exploring a suitable prediction model for the different sequence probabilities of the Chinese stock market has important reference value in quantifying timing,avoiding investment risks and making future investment decisions.This article uses a variety of time series methods to study the operating laws of the Chinese stock market,and uses three different types of data mining algorithms to predict the rise and fall of the Chinese stock market.In empirical research,first obtain the transaction data of all stocks in each trading day of the Chinese stock market from 2017 to 2019,and classify all sample data according to the increase to obtain the corresponding time series,and calculate the quantitative market sentiment and uncertainty advance-decline entropy index,and then conduct empirical research on the subdivided five types of advance-decline series and the corresponding advance-decline entropy series.The methods used in mining the rules of the advance-decline sequence of the Chinese stock market are mainly traditional time series methods.In the empirical study of the calendar effect,the autoregressive conditional heteroscedasticity(GARCH)model is used to analyze the monthly and weekly effects of the stock market rise series and the Shanghai stock index,In the periodic analysis of the advance-decline series,the spectral analysis method is mainly used.In the empirical research on the prediction of the different advance-decline series and the advance-decline entropy series in China's stock market,we use the traditional ARMA model of the prediction time series in the data mining algorithm,the support vector regression(SVR)algorithm in the machine learning algorithm,and the long short memory network(LSTM)algorithm in the deep learning to predict each advance-decline series.Finally,this paper comprehensively analyzes the empirical results obtained by the above methods and finds that the operation of the Chinese stock market has certain rules.The use of machine learning algorithms and deep learning algorithms in the data mining algorithm has relatively good results in predicting the advance-decline of the Chinese stock market.The study found that the advance-decline entropy index can reflect the uncertainty of the stock market.The month effect of the advance-decline entropy sequence corresponds to the Shanghai stock index and the daily limit sequence,and the major and secondary periods of the advance-decline entropy sequence are consistent with most of the graded series.This further confirms the uncertainty of the energy stock market volatility entropy index.In the past three years,there has been a significant February effect in the Chinese stock market.Among them,both the Shanghai stock index and the daily limit series have a significant positive February effect,while the rise and fall entropy series have a significant negative February effect.Only the daily limit and the Shanghai stock index have a week effect;The main periods of all series with gains greater than-5% and the rise and fall entropy series in each increase series are one month,and the main periods with all declines greater than 5% are shorter than the main cycles with all gains greater than-5%.Comprehensive analysis of the prediction results of three different types of data mining algorithms for predicting stock market ups and downs sequences.It is found that machine learning algorithms and deep learning algorithms are better than traditional time series models in fitting each series of gains and losses.Using LSTM in deep learning algorithms for China's stock market forecast the best results.