LT-SMF: long term stock market price trend prediction using optimal hybrid machine learning technique

Stock values are predicted using the market's biggest issues. Stock price data is hard to predict due to its distinct traits and volatility. Online news and comments reflect investor sentiment and thoughts on stocks, which may assist predict stock prices. Most authors lately recommended a stati...

Full description

Saved in:
Bibliographic Details
Published in:The Artificial intelligence review 2023-06, Vol.56 (6), p.5365-5402
Main Authors: Venkateswararao, K., Reddy, B. Venkata Ramana
Format: Article
Language:English
Subjects:
Accuracy
Artificial Intelligence
Classifiers
Computer Science
Digital media
Electronic news gathering
Error analysis
Forecasts and trends
Machine learning
Market prices
Mean square errors
Methods
Optimization
Predictions
Securities markets
Social networks
Stock markets
Stock prices
Stocks
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Stock values are predicted using the market's biggest issues. Stock price data is hard to predict due to its distinct traits and volatility. Online news and comments reflect investor sentiment and thoughts on stocks, which may assist predict stock prices. Most authors lately recommended a statistical and environmental measurement-based technique for anticipating machine learning or stock price changes. These models struggle with changing data. We propose a hybrid machine learning technique for long-term stock market price trend prediction (LT-SMF). We employed improved butterfly optimization (IBO) to remove artefacts from input data. Scaling, polarising, and variation percentage are used to find valuable qualities. Second, a brown Planthopper optimization (BPO) approach reduces data dimensionality for optimal feature selection. To forecast stock market price variations, a hybrid FEL-DNN was utilized. Using 11 stock market indices and social media data, evaluate the LT-SMF model. Simulation performance was compared to state-of-the-art models for mean square error, mean bios error, mean absolute error, root mean square error, accuracy, precision, recall, and F-measure. The proposed FEL-DNN classifier outperforms the current state-of-the-art CNN3D, CNN3D − DR, LSTM − D, CNN3D + LSTM, CNN3D − D + LSTM, and CNN3D − DR + LSTM classifiers by 38.67%, 41.71%, 39.32%, 36.04%, 41.13%, and 43.43% respectively in terms of accuracy in the social media data.
ISSN:0269-2821
1573-7462
DOI:10.1007/s10462-022-10291-5