To systematically incorporate multiple influencing factors,the coupled-state frequency memory(Co-SFM)network is proposed.This model integrates Copula estimation with neural networks,fusing multilevel data information,...To systematically incorporate multiple influencing factors,the coupled-state frequency memory(Co-SFM)network is proposed.This model integrates Copula estimation with neural networks,fusing multilevel data information,which is then fed into downstream learning modules.Co-SFM employs an upstream fusion module to incorporate multilevel data,thereby constructing a macro-plate-micro data structure.This configuration helps identify and integrate characteristics from different data levels,facilitating a deeper understanding of the internal links within the financial system.In the downstream model,Co-SFM uses a state-frequency memory network to mine hidden frequency information within stock prices,and the multifrequency patterns of sequential data are modeled.Empirical results show that Co-SFM s prediction accuracy for stock price trends is significantly better than that of other models.This is especially evident in multistep medium and long-term trend predictions,where integrating multilevel data results in notably improved accuracy.展开更多
Stock trend prediction is a challenging problem because it involves many variables.Aiming at the problem that some existing machine learning techniques, such as random forest(RF), probabilistic random forest(PRF), k-n...Stock trend prediction is a challenging problem because it involves many variables.Aiming at the problem that some existing machine learning techniques, such as random forest(RF), probabilistic random forest(PRF), k-nearest neighbor(KNN), and fuzzy KNN(FKNN), have difficulty in accurately predicting the stock trend(uptrend or downtrend) for a given date, a generalized Heronian mean(GHM) based FKNN predictor named GHM-FKNN was proposed.GHM-FKNN combines GHM aggregation function with the ideas of the classical FKNN approach.After evaluation, the comparison results elucidated that GHM-FKNN outperformed the other best existing methods RF, PRF, KNN and FKNN on independent test datasets corresponding to three stocks, namely AAPL, AMZN and NFLX.Compared with RF, PRF, KNN and FKNN, GHM-FKNN achieved the best performance with accuracy of 62.37% for AAPL, 58.25% for AMZN, and 64.10% for NFLX.展开更多
Stock market has a profound impact on the market economy,Hence,the prediction of future movement of stocks is of great significance to investors.Therefore,an efficient prediction system can solve this problem to a gre...Stock market has a profound impact on the market economy,Hence,the prediction of future movement of stocks is of great significance to investors.Therefore,an efficient prediction system can solve this problem to a great extent.In this paper,we used the stock price of Google Inc.as a prediction object,selected 3810 adjusted closing prices,and used long short-term memory(LSTM)method to predict the future price trend of the stock.We built a three-layer LSTM model and divided the entire data into a test set and a training set according to the ratio of 8 to 2.The final results show that while the LSTM model can predict the stock trend of Google Inc.very well,it cannot predict the specific price accurately.展开更多
Predicting stock price movement direction is a challenging problem influenced by different factors and capricious events. The conventional stock price prediction machine learning models heavily rely on the internal fi...Predicting stock price movement direction is a challenging problem influenced by different factors and capricious events. The conventional stock price prediction machine learning models heavily rely on the internal financial features, especially the stock price history. However, there are many outside-of-company features that deeply interact with the companies’ stock price performance, especially during the COVID period. In this study, we selected 9 COVID vaccine companies and collected their relevant features over the past 20 months. We added handcrafted external information, including COVID-related statistics and company-specific vaccine progress information. We implemented, evaluated, and compared several machine learning models, including Multilayer Perceptron Neural Networks with logistic regression and decision trees with boosting and bagging algorithms. The results suggest that the application of feature engineering and data mining techniques can effectively enhance the performance of models predicting stock price movement during the COVID period. The results show that COVID-related handcrafted features help to increase the model prediction accuracy by 7.3% and AUROC by 6.5% on average. Further exploration showed that with data selection the decision tree model with gradient, boosting algorithm achieved 70% in AUROC and 66% in the accuracy.展开更多
基金The National Natural Science Foundation of China(No.72173018).
文摘To systematically incorporate multiple influencing factors,the coupled-state frequency memory(Co-SFM)network is proposed.This model integrates Copula estimation with neural networks,fusing multilevel data information,which is then fed into downstream learning modules.Co-SFM employs an upstream fusion module to incorporate multilevel data,thereby constructing a macro-plate-micro data structure.This configuration helps identify and integrate characteristics from different data levels,facilitating a deeper understanding of the internal links within the financial system.In the downstream model,Co-SFM uses a state-frequency memory network to mine hidden frequency information within stock prices,and the multifrequency patterns of sequential data are modeled.Empirical results show that Co-SFM s prediction accuracy for stock price trends is significantly better than that of other models.This is especially evident in multistep medium and long-term trend predictions,where integrating multilevel data results in notably improved accuracy.
基金Supported by the National Key Research and Development Program (No.2019YFA0707201)the Key Work Program of Institute of Scientific and Technical Information of China (No.ZD2022-01,ZD2023-07)。
文摘Stock trend prediction is a challenging problem because it involves many variables.Aiming at the problem that some existing machine learning techniques, such as random forest(RF), probabilistic random forest(PRF), k-nearest neighbor(KNN), and fuzzy KNN(FKNN), have difficulty in accurately predicting the stock trend(uptrend or downtrend) for a given date, a generalized Heronian mean(GHM) based FKNN predictor named GHM-FKNN was proposed.GHM-FKNN combines GHM aggregation function with the ideas of the classical FKNN approach.After evaluation, the comparison results elucidated that GHM-FKNN outperformed the other best existing methods RF, PRF, KNN and FKNN on independent test datasets corresponding to three stocks, namely AAPL, AMZN and NFLX.Compared with RF, PRF, KNN and FKNN, GHM-FKNN achieved the best performance with accuracy of 62.37% for AAPL, 58.25% for AMZN, and 64.10% for NFLX.
文摘Stock market has a profound impact on the market economy,Hence,the prediction of future movement of stocks is of great significance to investors.Therefore,an efficient prediction system can solve this problem to a great extent.In this paper,we used the stock price of Google Inc.as a prediction object,selected 3810 adjusted closing prices,and used long short-term memory(LSTM)method to predict the future price trend of the stock.We built a three-layer LSTM model and divided the entire data into a test set and a training set according to the ratio of 8 to 2.The final results show that while the LSTM model can predict the stock trend of Google Inc.very well,it cannot predict the specific price accurately.
文摘Predicting stock price movement direction is a challenging problem influenced by different factors and capricious events. The conventional stock price prediction machine learning models heavily rely on the internal financial features, especially the stock price history. However, there are many outside-of-company features that deeply interact with the companies’ stock price performance, especially during the COVID period. In this study, we selected 9 COVID vaccine companies and collected their relevant features over the past 20 months. We added handcrafted external information, including COVID-related statistics and company-specific vaccine progress information. We implemented, evaluated, and compared several machine learning models, including Multilayer Perceptron Neural Networks with logistic regression and decision trees with boosting and bagging algorithms. The results suggest that the application of feature engineering and data mining techniques can effectively enhance the performance of models predicting stock price movement during the COVID period. The results show that COVID-related handcrafted features help to increase the model prediction accuracy by 7.3% and AUROC by 6.5% on average. Further exploration showed that with data selection the decision tree model with gradient, boosting algorithm achieved 70% in AUROC and 66% in the accuracy.
