Deep LSTM and LSTM-Attention Q-learning based reinforcement learning in oil and gas sector prediction

Knowledge-Based Systems, Journal Year: 2023, Volume and Issue: 284, P. 111290 - 111290

Published: Dec. 8, 2023

Accurate prediction of stock market trends and movements holds great significance in the financial industry as it enables investors, traders, decision-makers to make informed choices optimize their investment strategies. In context oil gas sector, where prices are influenced by complex dynamics various external factors, reliable predictions essential for effective decision-making risk management. This study proposes Deep Long Short-Term Memory Q-Learning (DLQL) Attention (DLAQL) models state-of-the-art (LSTM) predicting sector. The utilizes historical price data Cenovus Energy Inc. (CVE), MPLX LP (MPLX), Cheniere (LNG), Suncor (SU) create validate these models. research employs Markov Decision Process (MDP) framework, a widely-used reinforcement learning technique, train deep LSTM framework allows learn optimal policies based on data, enabling them accurate adapt changing conditions. findings this reveal that proposed DLQL DLAQL perform excellently well terms accuracy inclusion attention mechanisms model further enhances its performance allowing focus important features capture relevant information. results underscore potential within application can lead improved decision-making, enhanced management, increased profitability participants. Further exploration refinement models, along with incorporation additional variables indicators, contribute development more sophisticated future. Overall, contributes advancement techniques, specifically introducing evaluating efficacy highlight importance demonstrate benefits leveraging MDP support management dynamic competitive industry.

Language: Английский

---

Using Opinionated-Objective Terms to Improve Lexicon-Based Sentiment Analysis

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 23

Published: Jan. 1, 2024

Language: Английский

---

Graph-aware pre-trained language model for political sentiment analysis in Filipino social media

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 146, P. 110317 - 110317

Published: Feb. 20, 2025

Language: Английский

---

The Effectiveness of the Ensemble Naive Bayes in Analyzing Review Sentiment of the Lazada Application on Google Play

Published: Jan. 28, 2024

The surge in e-commerce growth Indonesia has led to the emergence of numerous new online marketplaces, including Lazada application. Within application, users encounter various experiences and can share their insights through reviews, highlighting both strengths weaknesses. However, abundance user reviews makes it challenging extract pertinent information tailored individual needs. Consequently, employing sentiment analysis becomes a viable solution sift review data, providing thorough assessment application's quality. methods used this research are Term Frequency - Inverse Document (TF-IDF) Ensemble Learning, specifically utilizing Voting approach. is compare effectiveness between single Naïve Bayes method. Multinomial demonstrates higher accuracy F1Score compared other classification models, achieving 89.1% 89.65% F1-Score.

Language: Английский

---

Deep Learning for Predicting Attrition Rate in Open and Distance Learning (ODL) Institutions

Computers, Journal Year: 2024, Volume and Issue: 13(9), P. 229 - 229

Published: Sept. 11, 2024

Student enrollment is a vital aspect of educational institutions, encompassing active, registered and graduate students. All the same, some students fail to engage with their studies after admission drop out along line; this known as attrition. The student attrition rate acknowledged most complicated significant problem facing systems caused by institutional non-institutional challenges. In study, researchers utilized dataset obtained from National Open University Nigeria (NOUN) 2012 2022, which included comprehensive information about enrolled in various programs at university who were inactive had dropped out. used deep learning techniques, such Long Short-Term Memory (LSTM) model compared performance One-Dimensional Convolutional Neural Network (1DCNN) model. results study revealed that LSTM achieved overall accuracy 57.29% on training data, while 1DCNN exhibited lower 49.91% data. indicated superior correct classification

Language: Английский

---

Enhancing Deep Learning-Based Sentiment Analysis Using Static and Contextual Language Models

Bitlis Eren Üniversitesi Fen Bilimleri Dergisi, Journal Year: 2023, Volume and Issue: 12(3), P. 712 - 724

Published: Sept. 21, 2023

Sentiment Analysis (SA) is an essential task of Natural Language Processing and used in various fields such as marketing, brand reputation control, social media monitoring. The scores generated by users product reviews are feedback sources for businesses to discover their products' positive or negative aspects. However, it takes work facing a large user population accurately assess the consistency scores. Recently, automated methodologies based on Deep Learning (DL), which utilize static especially pre-trained contextual language models, have shown successful performances SA tasks. To address issues mentioned above, this paper proposes Multi-layer Convolutional Neural Network-based approaches using Static Models (SLMs) Word2Vec GloVe Contextual (CLMs) ELMo BERT that can evaluate with ratings. Focusing improving model inputs sentence representations store richer features, study applied SLMs CLMs DL models evaluated impact performance. test performance proposed approaches, experimental studies were conducted Amazon dataset, publicly available considered benchmark dataset most researchers. According results studies, highest classification was obtained applying CLM 82% 84% training accuracy be domains' tasks provide insightful decision-making information.

Language: Английский

---

Political-RAG: using generative AI to extract political information from media content

Journal of Information Technology & Politics, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 16

Published: Oct. 23, 2024

In the digital era, media content is crucial for political analysis, providing valuable insights through news articles, social posts, speeches, and reports. Natural Language Processing (NLP) has transformed Political Information Extraction (IE), automating tasks such as event extraction sentiment analysis. Traditional NLP methods, while effective, are often task-specific require specialized expertise. contrast, Large Models (LLMs) powered by Generative Artificial Intelligence (GenAI) offer a more integrated solution. However, domain-specific challenges persist, which led to development of Retrieval-Augmented Generation (RAG) framework. RAG enhances LLMs incorporating external data retrieval, addressing issues related availability. To demonstrate RAG's capabilities, we introduce Political-RAG system, designed extract information from content, including Twitter articles. Initially developed extraction, system lays foundation developing various complex IE tasks. These include detecting hate speech, analyzing conflicts, assessing bias, evaluating trends, sentiment, opinions.

Language: Английский

---

SENTIMENT CLASSIFICATION OF TWEETS WITH EXPLICIT WORD NEGATIONS AND EMOJI USING DEEP LEARNING

International Journal of Computer Systems & Software Engineering, Journal Year: 2023, Volume and Issue: 9(2), P. 93 - 104

Published: July 20, 2023

The widespread use of social media platforms such as Twitter, Instagram, Facebook, and LinkedIn have had a huge impact on daily human interactions decision-making. Owing to Twitter's acceptance, users can express their opinions/sentiments nearly any issue, ranging from public opinion, product/service, even specific group people. Sharing these results in massive production user content known tweets, which be assessed generate new knowledge. Corporate insights, government policy formation, decision-making, brand identity monitoring all benefit analyzing the expressed tweets. Even though several techniques been created analyze sentiments engagements include negation words emoji elements that, if not properly pre-processed, would result misclassification. majority available pre-processing rely clean data machine learning algorithms annotate sentiment unlabeled texts. In this study, we propose text approach that takes into consideration characteristics by translating features single contextual tweets minimize context loss. proposed preprocessor was evaluated benchmark Twitter datasets using four deep algorithms: Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), Artificial (ANN). showed LSTM performed better than approaches already discussed literature, with an accuracy 96.36%, 88.41%, 95.39%. findings also suggest information like explicit word negations aids preservation sentimental information. This appears first study classify while accounting for both conversion translation.

Language: Английский

---

Analysis Sentiment towards Delivery Service: Case Study of Paxel

Published: Nov. 7, 2023

Paxel is an application-based delivery service. On Google Play, the application has been downloaded by more than 1 million users and 10 thousand reviews. From existing reviews, it will be useful for company if can processed properly. For this reason, a sentiment analysis was carried out to find sentiments of users, result used as reference improving services or products. The method in research Random Forest, Naive Bayes, Support Vector Machine (SVM). TF-IDF SMOTE data weighting unbalance. As well applying K-fold Cross Validation evaluate used. result, Forest higher accuracy value 91%, Bayes 83%, 87%. Where F1-Score on Positive Class 89%, Neutral 93% Negative 92%.

Language: Английский

---