When the country has changed with each passing day in the last several years,the stock market has changed a lot.The data size of the stock market is now very large and complex,and it is easily affected by many factors such as market conditions,domestic policies,and investor sentiment.The strong nature brings difficulty to the research of stock prediction.Multiple disciplines such as statistics,computer science,and machine learning haved formed data mining.It can search and mine effective information from huge data to support investors' decision-making and provide an effective way for stock data analysis.Data mining technology is very meaningful to study stock prediction.The paper first collected 27 technical index data of stocks and conducted index mining,and then established BP network and RBF network models to conduct predictive research and analysis on individual stocks such as Luo Niu Shan,Hainan Expressway and Shenzhen Stock Exchange Index.Insufficient optimization using genetic algorithm.Data mining technology has great application value in stock forecasting,and its innovative research results are as follows:(1)Whether the input variables for neural network predictive models are reasonable,the Apriori algorithm is proposed to analyze the correlation between the technical indicators of the stock,and determines the indicators related to the closing price for the next day.When the neural network model predicts Luonishan and Hainan high-speed stocks,its input variables are the highest price,the lowest price,the opening price,the closing price,the volume,the turnover,the turnover rate,MA1,MA2,MA3,and BOLL of the previous day,a total of 11 indicators.When predicting the broad market Shenzhen Stock Exchange Index,the input variables are the highest price,the lowest price,the opening price,the closing price,the turnover,MA1,MA2,MA3,BOLL,OBV,and RSI3 on the previous day.a total of 11 indicators.And compared with the input variables selected in most literatures(the price which is the highest,lowest,opening,closing,and the trading volume of the previous day),a simulation test was performed to compare the simulation results.The simulation results verified the feasibility of the indicators selected by the Apriori algorithm And effectiveness.(2)Aiming at the randomness and blindness of the trial and error method to determine the parameters of BP and RBF networks,an orthogonal test method is proposed to determine the parameters of the BP network(training samples,hidden layer)for predicting Luo Niushan,Hainan Expressway and Shenzhen Component Index.The optimal combination of the number of neurons,learning rate,learning times and expected error)and RBF network parameters(training samples,spread and expected error)were compared with the trial and error method.The simulation results were verified by the positive results The test method is used to determine the feasibility and effectiveness of BP and RBF network parameters.(3)Aiming at the problem that the orthogonal-BP network is easy to fall into local optimization,a genetic algorithm is proposed to optimize the orthogonal-BP network.And compared with orthogonal-BP network simulation test,the simulation results verified the feasibility and effectiveness of GA-orthogonal-BP network.The three methods and test results presented in this article are effective in improving the prediction accuracy of individual stocks and the broader market,and have application value,which can provide investors with meaningful technical support. |