Recurrent Neural Networks: A Comprehensive Review of Architectures, Variants, and Applications

Information, Год журнала: 2024, Номер 15(9), С. 517 - 517

Опубликована: Авг. 25, 2024

Recurrent neural networks (RNNs) have significantly advanced the field of machine learning (ML) by enabling effective processing sequential data. This paper provides a comprehensive review RNNs and their applications, highlighting advancements in architectures, such as long short-term memory (LSTM) networks, gated recurrent units (GRUs), bidirectional LSTM (BiLSTM), echo state (ESNs), peephole LSTM, stacked LSTM. The study examines application to different domains, including natural language (NLP), speech recognition, time series forecasting, autonomous vehicles, anomaly detection. Additionally, discusses recent innovations, integration attention mechanisms development hybrid models that combine with convolutional (CNNs) transformer architectures. aims provide ML researchers practitioners overview current future directions RNN research.

Язык: Английский

---

Recurrent Neural Networks: A Comprehensive Review of Architectures, Variants, and Applications

Опубликована: Авг. 12, 2024

Recurrent Neural Networks (RNNs) have significantly advanced the field of machine learning by enabling effective processing sequential data. This paper provides a comprehensive review RNNs and their applications, highlighting advancements in architectures such as Long Short-Term Memory (LSTM) networks, Gated Units (GRUs), Bidirectional LSTM (BiLSTM), stacked LSTM. The study examines application different domains, including natural language (NLP), speech recognition, financial time series forecasting, bioinformatics, autonomous vehicles, anomaly detection. Additionally, discusses recent innovations, integration attention mechanisms development hybrid models that combine with convolutional neural networks (CNNs) transformer architectures. aims to provide researchers practitioners overview current state future directions RNN research.

Язык: Английский

---

Machine Learning and Deep Learning Paradigms: From Techniques to Practical Applications and Research Frontiers

Computers, Год журнала: 2025, Номер 14(3), С. 93 - 93

Опубликована: Март 6, 2025

Machine learning (ML) and deep (DL), subsets of artificial intelligence (AI), are the core technologies that lead significant transformation innovation in various industries by integrating AI-driven solutions. Understanding ML DL is essential to logically analyse applicability identify their effectiveness different areas like healthcare, finance, agriculture, manufacturing, transportation. consists supervised, unsupervised, semi-supervised, reinforcement techniques. On other hand, DL, a subfield ML, comprising neural networks (NNs), can deal with complicated datasets health, autonomous systems, finance industries. This study presents holistic view technologies, analysing algorithms application’s capacity address real-world problems. The investigates application which techniques implemented. Moreover, highlights latest trends possible future avenues for research development (R&D), consist developing hybrid models, generative AI, incorporating technologies. aims provide comprehensive on serve as reference guide researchers, industry professionals, practitioners, policy makers.

Язык: Английский

---

RNA structure prediction using deep learning — A comprehensive review

Computers in Biology and Medicine, Год журнала: 2025, Номер 188, С. 109845 - 109845

Опубликована: Фев. 20, 2025

In computational biology, accurate RNA structure prediction offers several benefits, including facilitating a better understanding of functions and RNA-based drug design. Implementing deep learning techniques for has led tremendous progress in this field, resulting significant improvements accuracy. This comprehensive review aims to provide an overview the diverse strategies employed predicting secondary structures, emphasizing methods. The article categorizes discussion into three main dimensions: feature extraction methods, existing state-of-the-art model architectures, approaches. We present comparative analysis various models highlighting their strengths weaknesses. Finally, we identify gaps literature, discuss current challenges, suggest future approaches enhance performance applicability tasks. provides deeper insight subject paves way further dynamic intersection life sciences artificial intelligence.

Язык: Английский

---

Generative Modeling for Imbalanced Credit Card Fraud Transaction Detection

Journal of Cybersecurity and Privacy, Год журнала: 2025, Номер 5(1), С. 9 - 9

Опубликована: Март 17, 2025

The increasing sophistication of fraud tactics necessitates advanced detection methods to protect financial assets and maintain system integrity. Various approaches based on artificial intelligence have been proposed identify fraudulent activities, leveraging techniques such as machine learning deep learning. However, class imbalance remains a significant challenge. We propose several solutions generative modeling address the challenges posed by in detection. Class often hinders performance models limiting their ability learn from minority classes, transactions. Generative offer promising approach mitigate this issue creating realistic synthetic samples, thereby enhancing model’s detect rare cases. In study, we introduce evaluate multiple models, including Variational Autoencoders (VAEs), standard (AEs), Adversarial Networks (GANs), hybrid Autoencoder–GAN model (AE-GAN). These aim generate samples balance dataset improve capacity. Our primary objective is compare these against traditional oversampling techniques, SMOTE ADASYN, context conducted extensive experiments using real-world credit card effectiveness our solutions. results, measured BEFS metrics, demonstrate that not only problem more effectively but also outperform conventional identifying

Язык: Английский

---

A Hybrid Deep Learning Approach with Generative Adversarial Network for Credit Card Fraud Detection

Technologies, Год журнала: 2024, Номер 12(10), С. 186 - 186

Опубликована: Окт. 2, 2024

Credit card fraud detection is a critical challenge in the financial industry, with substantial economic implications. Conventional machine learning (ML) techniques often fail to adapt evolving patterns and underperform imbalanced datasets. This study proposes hybrid deep framework that integrates Generative Adversarial Networks (GANs) Recurrent Neural (RNNs) enhance capabilities. The GAN component generates realistic synthetic fraudulent transactions, addressing data imbalance enhancing training set. discriminator, implemented using various DL architectures, including Simple RNN, Long Short-Term Memory (LSTM) networks, Gated Units (GRUs), trained distinguish between real transactions further fine-tuned classify as or legitimate. Experimental results demonstrate significant improvements over traditional methods, GAN-GRU model achieving sensitivity of 0.992 specificity 1.000 on European credit dataset. work highlights potential GANs combined architectures provide more effective adaptable solution for detection.

Язык: Английский

---

Deep Learning in Finance: A Survey of Applications and Techniques

AI, Год журнала: 2024, Номер 5(4), С. 2066 - 2091

Опубликована: Окт. 28, 2024

Machine learning (ML) has transformed the financial industry by enabling advanced applications such as credit scoring, fraud detection, and market forecasting. At core of this transformation is deep (DL), a subset ML that robust in processing analyzing complex large datasets. This paper provides comprehensive overview key models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Deep Belief (DBNs), Transformers, Generative Adversarial (GANs), Reinforcement Learning (Deep RL). Beyond summarizing their mathematical foundations processes, study offers new insights into how these models are applied real-world contexts, highlighting specific advantages limitations tasks algorithmic trading, risk management, portfolio optimization. It also examines recent advances emerging trends alongside critical challenges data quality, model interpretability, computational complexity. These can guide future research directions toward developing more efficient, robust, explainable address evolving needs sector.

Язык: Английский

---

A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications

Information, Год журнала: 2024, Номер 15(12), С. 755 - 755

Опубликована: Ноя. 27, 2024

Deep learning (DL) has become a core component of modern artificial intelligence (AI), driving significant advancements across diverse fields by facilitating the analysis complex systems, from protein folding in biology to molecular discovery chemistry and particle interactions physics. However, field deep is constantly evolving, with recent innovations both architectures applications. Therefore, this paper provides comprehensive review DL advances, covering evolution applications foundational models like convolutional neural networks (CNNs) Recurrent Neural Networks (RNNs), as well such transformers, generative adversarial (GANs), capsule networks, graph (GNNs). Additionally, discusses novel training techniques, including self-supervised learning, federated reinforcement which further enhance capabilities models. By synthesizing developments identifying current challenges, insights into state art future directions research, offering valuable guidance for researchers industry experts.

Язык: Английский

---

Deep BiLSTM Attention Model for Spatial and Temporal Anomaly Detection in Video Surveillance

Sensors, Год журнала: 2025, Номер 25(1), С. 251 - 251

Опубликована: Янв. 4, 2025

Detection of anomalies in video surveillance plays a key role ensuring the safety and security public spaces. The number cameras is growing, making it harder to monitor them manually. So, automated systems are needed. This change increases demand for that detect abnormal events or anomalies, such as road accidents, fighting, snatching, car fires, explosions real-time. These improve detection accuracy, minimize human error, make operations more efficient. In this study, we proposed Composite Recurrent Bi-Attention (CRBA) model detecting videos. CRBA combines DenseNet201 robust spatial feature extraction with BiLSTM networks capture temporal dependencies across frames. A multi-attention mechanism was also incorporated direct model’s focus critical spatiotemporal regions. improves system’s ability distinguish between normal behaviors. By integrating these methodologies, classification videos, effectively addressing both challenges. Experimental assessments demonstrate achieves high accuracy on University Central Florida (UCF) newly developed Road Anomaly Dataset (RAD). enhances while improving resource efficiency minimizing response times situations. advantages an invaluable tool operations, where rapid accurate responses needed maintaining safety.

Язык: Английский

---

Achieving Excellence in Cyber Fraud Detection: A Hybrid ML+DL Ensemble Approach for Credit Cards

Applied Sciences, Год журнала: 2025, Номер 15(3), С. 1081 - 1081

Опубликована: Янв. 22, 2025

The rapid advancement of technology has increased the complexity cyber fraud, presenting a growing challenge for banking sector to efficiently detect fraudulent credit card transactions. Conventional detection approaches face challenges in adapting continuously evolving tactics fraudsters. This study addresses these limitations by proposing an innovative hybrid model that integrates Machine Learning (ML) and Deep (DL) techniques through stacking ensemble resampling strategies. leverages ML including Decision Tree (DT), Random Forest (RF), Support Vector (SVM), eXtreme Gradient Boosting (XGBoost), Categorical (CatBoost), Logistic Regression (LR) alongside DL such as Convolutional Neural Network (CNN) Bidirectional Long Short-Term Memory (BiLSTM) with attention mechanisms. By utilising method, consolidates predictions from multiple base models, resulting improved predictive accuracy compared individual models. methodology incorporates robust data pre-processing techniques. Experimental evaluations demonstrate superior performance ML+DL model, particularly handling class imbalances achieving high F1 score, score 94.63%. result underscores effectiveness proposed delivering reliable fraud detection, highlighting its potential enhance financial transaction security.

Язык: Английский

---