Research On Prediction Of Public Opinion In Stock Market Network Based On AdaBoost-IWOA-Elman Algorithm

The performance of the stock market is often the "barometer" of a country's economic development,while the investors in china's stock market are mainly retail investors,relying on news reports,analysis reports or some small news to analyze,buy and sell stocks.In recent years,with the development of network technology,more and more shareholders like to publish or obtain relevant stock comments through the network platform as stock investment channel,which is called stock market online public opinion.The stock market online public opinion mostly exists in the form of stock comments in the stock forum,about 90% of which is text data,which has the characteristics of massive,unstructured and real-time changes.Therefore,how to use the fragment comment information on the network to predict the development trend of the stock market,so as to optimize their investment decision-making is a hot and difficult research topic at present.In this paper,the comment text of Shanghai Stock Exchange(SSE)180 index stock shares of Oriental Wealth and the closing price in the corresponding Wind database in 2016 are selected as the subject investigated.Based on the Complete Ensemble Empirical Mode Decomposition With Adaptive Noise(CEEMDAN)algorithm,the paper constructs AdaBoost-IWOA-Elman prediction model through text mining technology,which provides a feasible prediction method for the closing price prediction based on the online public opinion of the stock market.The specific work is as follows:(1)Acquisition and quantification of text data.Python Spyder web crawler is used to crawl the related information of stock market comments on Oriental Wealth and storage it in the form of.CSV.Python Jieba algorithm is used to segment the comment information.Then use the reconstructed stoppage list to remove the high-frequency and meaningless stop words.At the same time,combine synonymous words in the data.Then the Term Frequency–Inverse Document Frequency(TF-IDF)algorithm is used to calculate the importance of the public opinion feature words in the overall public opinion data corpus.Finally,Vector Space Model(VSM)Model is used to represent the quantified data set according to the weight of the feature words(2)Important attribute selection.In this paper,the Boruta algorithm is used to filter the attributes to reduce the complexity of the attribute set,and then the CEEMDAN algorithm is used to realize the decomposition and noise reduction of the attribute sequence by adding a certain amount of white noise with specific variance to each attribute.Finally,the similar Intrinsic Mode Function(IMF)component and R remainder of the decomposed attribute values are combined,and the combined attribute set is used as the final modeling attribute set.(3)Build AdaBoost-IWOA-Elman prediction model.Firstly,Whale Optimization Algorithm(WOA)algorithm is used to optimize the initial weight and threshold value of Elman neural network in the iterative process,which effectively solves the problem of low prediction accuracy of Elman neural network.At the same time,in order to enhance the global and local search ability of WOA algorithm,adaptive weight is used to improve WOA algorithm,which makes WOA algorithm has appropriate nonlinear weight in the iterative process.Finally,five weak IWOA-Elman predictors are combined into a strong predictor by AdaBoost algorithm in continuous iteration,thus improving the prediction accuracy.The results show that,the proposed model is better than Elman neural network,its mean absolute error(MAE)reduces from 358.8120 to 113.0553,comparing with the original data set without CEEMDAN algorithm,its mean absolute error(MAPE)reduces from 4.9423% to 1.44531%,which effectively improves the prediction accuracy and provides an effective experimental method for forecasting the public opinion of stock market network.