|
Deep Learning in Sentiment Analysis of Investment Smart Networks Users
|
Nasrin Salehi Chegeni   , Saba Joudaki , Mojtaba Salehi |
| IAU, Khorramabad |
|
|
Abstract: (277 Views) |
Analyzing company stock data is one of the important methods for evaluating the performance and value of companies and choosing the best option for investing in financial markets. In this article, the data collection of the opinions of the users of the smart investment network of stock finder has been collected with the aim of analyzing the sentiments of the users. First decision tree, support vector machine, simple Bayesian and nearest neighbor algorithms were implemented; that the support vector achieved the best performance with 61 % accuracy. Then to compare these traditional learning algorithms with deep learning algorithms, LSTM and BERT networks were implemented in Farsi and BERT in English. These models performed better than traditional algorithms with accuracy of 72, 82 and 83 % respectively.Next the LSTM model was implemented using the meta-heuristic algorithm of genetics with the aim of obtaining optimal hyper-parameters, which reached an accuracy of 81.46 %. In the final phase, the bilingual BRET algorithm was implemented by combining the Persian text of comments and their English meaning and achieved a performance of 84 %. It is expected that the use of this model can help improve the performance of predictive and recommender systems in economic sites.
|
Article number: 3 |
| Keywords: Natural Language Processing, Sentiment Analysis, Deep Learning, Neural Network, Recommender Systems |
|
|
Full-Text [PDF 1648 kb]
(114 Downloads)
|
Type of Study: Research |
Subject:
Special
Received: 2024/10/7 | Accepted: 2025/02/7 | Published: 2025/02/18
|
|
|
|
|
|
|
| Add your comments about this article |
|
|
|
