{"title":"基于注意机制的双尺度自适应剩余长短期记忆的创新情绪对股市预测的影响","authors":"R. Gnanavel, J. M. Gnanasekar","doi":"10.1111/coin.70073","DOIUrl":null,"url":null,"abstract":"<div>\n \n <p>The stock market is extremely unpredictable and impulsive because of a variety of reasons, including public opinion, economic conditions, and so on. Each second, many Petabytes of data emerge from various sources, impacting the stock marketplace. A fair and effective merging of those sources of information (factors) into knowledge is predicted to improve the precision of stock market predictions. However, combining these characteristics from multiple sources of data into a single dataset to supply market evaluation is considered difficult since they are presented in various formats. This paper recommends a deep learning framework for performing prediction in the stock market by considering the sentiment text and historical information from social media. Initially, the required sentiment text and data are collected from the social media platform. From the database, the historical data of the company and the sentiment text from the user uploaded in the social media and news articles are collected. After that, the collected sentiment texts are preprocessed to remove the unwanted data. The preprocessed sentiment texts are given to the Bidirectional Encoder Representations from Transformers (BERT) model for retrieving the first set of features from the positive and negative sentiments. On the other hand, the deep features are retrieved from the data using a One-Dimensional Convolutional Neural Network (1DCNN), which is considered a second feature set from historical data. The two sets of features retrieved from the sentiment text and data are passed to the Dual Scale Adaptive Residual Long Short-Term Memory with Attention Mechanism (DSAResLSTM-AM) for stock market price prediction, where the attributes of the ResLSTM are tuned using Enhanced Deep Sleep Optimizer (EDSO). Here, the sentiment text having positive and negative sentiments helps to predict the stock market price of the company effectively to be less or high along with the analysis of previous data. The recommended model helps to perform the accurate stock market prediction, and it is used to enhance the return and reduce the investment. Finally, experimental validations are conducted to find the performance of the developed model in the stock market prediction.</p>\n </div>","PeriodicalId":55228,"journal":{"name":"Computational Intelligence","volume":"41 3","pages":""},"PeriodicalIF":1.8000,"publicationDate":"2025-06-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"An Innovative Sentiment Influenced Stock Market Prediction Based on Dual Scale Adaptive Residual Long Short Term Memory With Attention Mechanism\",\"authors\":\"R. Gnanavel, J. M. Gnanasekar\",\"doi\":\"10.1111/coin.70073\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div>\\n \\n <p>The stock market is extremely unpredictable and impulsive because of a variety of reasons, including public opinion, economic conditions, and so on. Each second, many Petabytes of data emerge from various sources, impacting the stock marketplace. A fair and effective merging of those sources of information (factors) into knowledge is predicted to improve the precision of stock market predictions. However, combining these characteristics from multiple sources of data into a single dataset to supply market evaluation is considered difficult since they are presented in various formats. This paper recommends a deep learning framework for performing prediction in the stock market by considering the sentiment text and historical information from social media. Initially, the required sentiment text and data are collected from the social media platform. From the database, the historical data of the company and the sentiment text from the user uploaded in the social media and news articles are collected. After that, the collected sentiment texts are preprocessed to remove the unwanted data. The preprocessed sentiment texts are given to the Bidirectional Encoder Representations from Transformers (BERT) model for retrieving the first set of features from the positive and negative sentiments. On the other hand, the deep features are retrieved from the data using a One-Dimensional Convolutional Neural Network (1DCNN), which is considered a second feature set from historical data. The two sets of features retrieved from the sentiment text and data are passed to the Dual Scale Adaptive Residual Long Short-Term Memory with Attention Mechanism (DSAResLSTM-AM) for stock market price prediction, where the attributes of the ResLSTM are tuned using Enhanced Deep Sleep Optimizer (EDSO). Here, the sentiment text having positive and negative sentiments helps to predict the stock market price of the company effectively to be less or high along with the analysis of previous data. The recommended model helps to perform the accurate stock market prediction, and it is used to enhance the return and reduce the investment. Finally, experimental validations are conducted to find the performance of the developed model in the stock market prediction.</p>\\n </div>\",\"PeriodicalId\":55228,\"journal\":{\"name\":\"Computational Intelligence\",\"volume\":\"41 3\",\"pages\":\"\"},\"PeriodicalIF\":1.8000,\"publicationDate\":\"2025-06-07\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Computational Intelligence\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1111/coin.70073\",\"RegionNum\":4,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computational Intelligence","FirstCategoryId":"94","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1111/coin.70073","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
An Innovative Sentiment Influenced Stock Market Prediction Based on Dual Scale Adaptive Residual Long Short Term Memory With Attention Mechanism
The stock market is extremely unpredictable and impulsive because of a variety of reasons, including public opinion, economic conditions, and so on. Each second, many Petabytes of data emerge from various sources, impacting the stock marketplace. A fair and effective merging of those sources of information (factors) into knowledge is predicted to improve the precision of stock market predictions. However, combining these characteristics from multiple sources of data into a single dataset to supply market evaluation is considered difficult since they are presented in various formats. This paper recommends a deep learning framework for performing prediction in the stock market by considering the sentiment text and historical information from social media. Initially, the required sentiment text and data are collected from the social media platform. From the database, the historical data of the company and the sentiment text from the user uploaded in the social media and news articles are collected. After that, the collected sentiment texts are preprocessed to remove the unwanted data. The preprocessed sentiment texts are given to the Bidirectional Encoder Representations from Transformers (BERT) model for retrieving the first set of features from the positive and negative sentiments. On the other hand, the deep features are retrieved from the data using a One-Dimensional Convolutional Neural Network (1DCNN), which is considered a second feature set from historical data. The two sets of features retrieved from the sentiment text and data are passed to the Dual Scale Adaptive Residual Long Short-Term Memory with Attention Mechanism (DSAResLSTM-AM) for stock market price prediction, where the attributes of the ResLSTM are tuned using Enhanced Deep Sleep Optimizer (EDSO). Here, the sentiment text having positive and negative sentiments helps to predict the stock market price of the company effectively to be less or high along with the analysis of previous data. The recommended model helps to perform the accurate stock market prediction, and it is used to enhance the return and reduce the investment. Finally, experimental validations are conducted to find the performance of the developed model in the stock market prediction.
期刊介绍:
This leading international journal promotes and stimulates research in the field of artificial intelligence (AI). Covering a wide range of issues - from the tools and languages of AI to its philosophical implications - Computational Intelligence provides a vigorous forum for the publication of both experimental and theoretical research, as well as surveys and impact studies. The journal is designed to meet the needs of a wide range of AI workers in academic and industrial research.