RNN-LSTM: From applications to modeling techniques and beyond—Systematic review

Journal of King Saud University - Computer and Information Sciences, Journal Year: 2024, Volume and Issue: 36(5), P. 102068 - 102068

Published: May 21, 2024

Long Short-Term Memory (LSTM) is a popular Recurrent Neural Network (RNN) algorithm known for its ability to effectively analyze and process sequential data with long-term dependencies. Despite popularity, the challenge of initializing optimizing RNN-LSTM models persists, often hindering their performance accuracy. This study presents systematic literature review (SLR) using an in-depth four-step approach based on PRISMA methodology, incorporating peer-reviewed articles spanning 2018-2023. It aims address how weight initialization optimization techniques can bolster performance. SLR offers detailed overview across various applications domains, stands out by comprehensively analyzing modeling techniques, datasets, evaluation metrics, programming languages associated networks. The findings this provide roadmap researchers practitioners enhance networks achieve superior results.

Language: Английский

---

A survey of Beluga whale optimization and its variants: Statistical analysis, advances, and structural reviewing

Computer Science Review, Journal Year: 2025, Volume and Issue: 57, P. 100740 - 100740

Published: March 3, 2025

Language: Английский

---

Transforming Smart Homes Via P2P Energy Trading Using Robust Forecasting and Scheduling Framework

Results in Engineering, Journal Year: 2024, Volume and Issue: 23, P. 102766 - 102766

Published: Aug. 24, 2024

With the advent of smart grids, advanced information infrastructures, metering facilities, bidirectional exchange information, and battery storage home area networks have all transformed electricity consumption energy efficiency. There is a significant shift in management structure from traditional centralized infrastructure to flexible demand side driven cyber-physical power systems with clean system. These changes significantly evolved (HEM) space. Consequently, stakeholders must define their responsibilities, create efficient regulatory frameworks, test out novel commercial strategies. P2P trading appears be feasible solution these circumstances, allowing users trade one another before becoming completely reliant on utility. offers more stable platform for by facilitating between prosumers consumers. This research proposes generation prediction techniques HEMS optimal using Multi-Objective Optimization model. An enhanced Wild Horse technique was first used summarize historical records' qualities. Then, Bi-LSTM predict values. Furthermore, Grasshopper optimization (GHO) approach employed fine-tune model's hyperparameters. The framework offered probabilistic fault evaluation that upholds load flow balance need supply continuous operations. It results an intelligent community transforming cities into ones, opening new avenues scientific terms technological developments.

Language: Английский

---

Flood susceptibility assessment using deep neural networks and open-source spatial datasets in transboundary river basin

VIETNAM JOURNAL OF EARTH SCIENCES, Journal Year: 2025, Volume and Issue: unknown

Published: April 16, 2025

The Mekong Basin is the most critical transboundary river basin in Asia. This provides an abundant source of fresh water essential for development agriculture, domestic consumption, and industry, as well production hydroelectricity, it also contributes to ensuring food security worldwide. region often subject floods that cause significant damage human life, society, economy. However, flood risk management challenges this are increasingly substantial due conflicting objectives between several countries data sharing. study integrates deep learning with optimization algorithms, namely Grasshopper Optimisation Algorithm (GOA), Adam Stochastic Gradient Descent (SGD), open-source datasets identify probably occurring basin, covering Vietnam Cambodia. Various statistical indices, Area Under Curve (AUC), root mean square error (RMSE), absolute (MAE), coefficient determination (R²), were used evaluate susceptibility models. results show proposed models performed AUC values above 0.8, specifying DNN-Adam model achieved 0.98, outperforming DNN-GOA (AUC = 0.89), DNN-SGD 0.87), XGB 0.82. Regions very high concentrated Delta along River findings supporting decision-makers or planners proposing appropriate mitigation strategies, planning policies, particularly watershed.

Language: Английский

---

Multi-strategy ensemble binary RIME optimization for feature selection

Neurocomputing, Journal Year: 2025, Volume and Issue: 639, P. 130284 - 130284

Published: April 19, 2025

Language: Английский

---

A novel conditional discrimination index approach for feature selection in partially labeled hybrid data

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(7)

Published: May 15, 2025

Language: Английский

---

Design of an Adaptive Robust PI Controller for DC/DC Boost Converter Using Reinforcement-Learning Technique and Snake Optimization Algorithm

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 141814 - 141829

Published: Jan. 1, 2024

The DC/DC Boost converter exhibits a non-minimum phase system with right half-plane zero structure, posing significant challenges for the design of effective control approaches. This article presents robust Proportional-Integral (PI) controller this an online adaptive mechanism based on Reinforcement-Learning (RL) strategy. Classical PI controllers are simple and easy to build, but they need be more against wide range disturbances adaptable operational parameters. To address these issues, RL strategy is used optimize performance controller. Some main advantages lower sensitivity error, reliable results through collection data from environment, ideal model behavior within specific context, better frequency matching in real-time applications. Random exploration, nevertheless, can result disastrous outcomes surprising real-world settings. Therefore, we opt Deterministic Policy Gradient (DPG) technique, which employs deterministic action function as opposed stochastic one. DPG combines benefits actor-critics, deep Q-networks, policy gradient method. In addition, method adopts Snake Optimization (SO) algorithm initial condition gains, yielding faster dynamics. SO known its disciplined nature-inspired approach, decision-making greater accuracy compared other optimization algorithms. A structure using hardware setup CONTROLLINO MAXI Automation built, offers cost-effective precise measurement Finally, achieved by simulations experiments demonstrate robustness approach.

Language: Английский

---

Hierarchical Learning-Enhanced Chaotic Crayfish Optimization Algorithm: Improving Extreme Learning Machine Diagnostics in Breast Cancer

Mathematics, Journal Year: 2024, Volume and Issue: 12(17), P. 2641 - 2641

Published: Aug. 26, 2024

Extreme learning machines (ELMs), single hidden-layer feedforward neural networks, are renowned for their speed and efficiency in classification regression tasks. However, generalization ability is often undermined by the random generation of hidden layer weights biases. To address this issue, paper introduces a Hierarchical Learning-based Chaotic Crayfish Optimization Algorithm (HLCCOA) aimed at enhancing ELMs. Initially, to resolve problems slow search premature convergence typical traditional crayfish optimization algorithms (COAs), HLCCOA utilizes chaotic sequences population position initialization. The ergodicity chaos leveraged boost diversity, laying groundwork effective global efforts. Additionally, hierarchical mechanism encourages under-performing individuals engage extensive cross-layer enhanced exploration, while top performers directly learn from elite highest improve local exploitation abilities. Rigorous testing with CEC2019 CEC2022 suites shows HLCCOA’s superiority over both original COA nine heuristic algorithms. Ultimately, HLCCOA-optimized extreme machine model, HLCCOA-ELM, exhibits superior performance reported benchmark models terms accuracy, sensitivity, specificity UCI breast cancer diagnosis, underscoring practicality robustness, as well HLCCOA-ELM’s commendable performance.

Language: Английский

---

A multipopulation grasshopper algorithm for multiobjective material distribution scheduling problem of assembly line with flexible line inventory

Robotic Intelligence and Automation, Journal Year: 2024, Volume and Issue: 44(5), P. 724 - 745

Published: Aug. 22, 2024

Purpose Owing to the finite nature of boundary line (BOL), conventional method, involving strong matching single-variety parts with storage locations at periphery line, proves insufficient for mixed-model assembly lines (MMAL). Consequently, this paper aims introduce a material distribution scheduling problem considering shared area (MDSPSSA). To address inherent trade-off requirement achieving both just-in-time efficiency and energy savings, mathematical model is developed bi-objectives minimizing line-side inventory consumption. Design/methodology/approach A nondominated multipopulation multiobjective grasshopper optimization algorithm (NM-MOGOA) proposed medium-to-large-scale associated MDSPSSA. This combines elements from sorting genetic algorithm-II. The coevolutionary strategy, chaotic mapping two further operators are used enhance overall solution quality. Findings Finally, performance evaluated by comparing NM-MOGOA multi-objective grey wolf optimizer, equilibrium optimizer atomic orbital search. experimental findings substantiate efficacy NM-MOGOA, demonstrating its promise as robust when confronted challenges posed MDSPSSA in MMALs. Originality/value system devised takes into account establishment areas between adjacent workstations. It permits undifferentiated various part types fixed BOL areas. Concurrently, innovative serves core system, supporting formulation plans.

Language: Английский

---

An Efficient Malware Classification Method Using Novel Deep Learning Model

AURUM Journal of Engineering Systems and Architecture, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 20, 2024

Malware attacks getting increased due to the complexity in their structures have become a key threat cybersecurity and require better more efficient means of detection. Signature heuristic methods detecting malware do not perform well slow developments this field thus current detection uses machine learning deep approaches. However, it is seen that high dimensionality data are major problems terms existing solutions, such as computational burden overfitting. The presented work thesis aims design new framework using ResNet50 neural networks fine-tuned with wrapper-based feature selection technique operated by GOA. supporting also takes advantage transfer ResNet50, robust convolutional network, for extraction from data. Every slight hint related learnt model through training on datasets. In addition this, GOA-based approach used help define most important features input network relieve load. To assess effectiveness proposed approach, benchmark datasets were used, results compared traditional recent methods. findings affirm ResNet50-GOA fine-tuning outperforms competitors significant margin rate improved accuracy, precision, recall, area under precision-recall curve, F1-score, which illustrates robustness fewer false positive cases complex computation. addition, immune issues like class imbalance discovers patterns emerging malware. This paper fulfills following gaps literature: It proposes scalable than metaheuristic optimization algorithms. speak promise combination techniques addressing multi-faceted issues, opens further possibilities improvement automated identification systems future

Language: Английский

---