Greenhouse Environment Sentinel with Hybrid LSTM-SVM for Proactive Climate Management

AgriEngineering, Journal Year: 2025, Volume and Issue: 7(4), P. 96 - 96

Published: April 1, 2025

This research presents a hybrid approach of Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) model for greenhouse environmental monitoring, integrating machine learning Internet Things (IoT)-based sensing to enhance climate prediction classification. Unlike traditional single-method approaches, this dual-model system provides comprehensive framework real-time control, optimizing temperature humidity forecasting while enabling accurate weather The LSTM excels in capturing sequential patterns, achieving superior performance with Root-Mean-Square Error (RMSE) 0.0766, Mean Absolute (MAE) 0.0454, coefficient determination (R2) 0.8825. For forecasting, our comparative analysis revealed that the Simple Recurrent Neural Network (RNN) demonstrates best accuracy (RMSE: 5.3034, MAE: 3.8041, R2: 0.8187), an unexpected finding highlights importance parameter-specific selection. Simultaneously, SVM classifies states 0.63, surpassing classifiers such as Logistic Regression K Nearest Neighbors (KNN). To data collection transmission, ESP NOW wireless protocol is integrated, ensuring low latency reliable communication between sensors. proposed LSTM-SVM system, combined IoT technology, represents significant advancement proactive management, offering scalable sustainable solution plant growth, resource allocation, adaptation.

Language: Английский

---

Machine Learning in Maritime Safety for Autonomous Shipping: A Bibliometric Review and Future Trends

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(4), P. 746 - 746

Published: April 8, 2025

Autonomous vessels are becoming paramount to ocean transportation, while they also face complex risks in dynamic marine environments. Machine learning plays a crucial role enhancing maritime safety by leveraging its data analysis and predictive capabilities. However, there has been no review grounded bibliometric this field. To explore the research evolution knowledge frontier field of for autonomous shipping, was conducted using 719 publications from Web Science database, covering period 2000 up May 2024. This study utilized VOSviewer, alongside traditional literature methods, construct network map perform cluster analysis, thereby identifying hotspots, trends, emerging frontiers. The findings reveal robust cooperative among journals, researchers, institutions, countries or regions, underscoring interdisciplinary nature domain. Through review, we found that machine methods evolving toward systematic comprehensive direction, integration with AI human interaction may be next bellwether. Future will concentrate on three main areas: objectives towards proactive management coordination, developing advanced technologies, such as bio-inspired sensors, quantum learning, self-healing systems, decision-making algorithms generative adversarial networks (GANs), hierarchical reinforcement (HRL), federated learning. By visualizing collaborative networks, analyzing evolutionary lays groundwork pioneering advancements sets visionary angle future shipping. Moreover, it facilitates partnerships between industry academia, making concerted efforts domain USVs.

Language: Английский

---

Predictive Analysis of Maritime Congestion Using Dynamic Big Data and Multiscale Feature Analysis

Journal of Electrical and Computer Engineering, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

The maritime industry is one of the most crucial sectors in global economy, facilitating transportation goods and commodities across vast distances. However, network congestion has become an increasingly critical challenge that significantly affects shipping efficiency overall operational performance industry. This study proposes innovative prediction approach using dynamic big data analysis vessel trajectories multiscale feature analysis. First, aims to extract valuable information from ships’ as they navigate oceans, enabling proactive traffic management optimized routing. Second, provides a comprehensive understanding by examining it different perspectives scales, leading more accurate predictions effective strategies. Furthermore, this introduces enhanced Faster R‐CNN detection model for real‐time tracking, integrating convolutional SKNet networks. To improve short‐term flow accuracy, employs through wavelet transformation. foundational undergo decomposition detailed representation frequencies. Gated recurrent unit (GRU) neural autoregressive moving average (ARMA) models are utilized predict trend noise components, respectively. Fusion demonstrates superior accuracy validated against real data. Empirical results showcase minimal errors heightened compared actual

Language: Английский

---

All-day perception for intelligent vehicles: switching perception algorithms based on WBCNet

Science China Information Sciences, Journal Year: 2024, Volume and Issue: 67(11)

Published: Oct. 22, 2024

Language: Английский

---

A novel deep learning model for predicting marine pollution for sustainable ocean management

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2482 - e2482

Published: Nov. 25, 2024

Climate change has become a major source of concern to the global community. The steady pollution environment including our waters is gradually increasing effects climate change. disposal plastics in seas alters aquatic life. Marine plastic poses grave danger marine and long-term health ocean. Though technology also seen as one contributors many aspects it are being applied combat climate-related disasters raise awareness about need protect planet. This study investigated amount undersea leveraging power artificial intelligence identify categorise wastes. classification was done using two types machine learning algorithms: two-step clustering fully convolutional network (FCN). models were trained Kaggle’s location data, which acquired situ . An experimental test conducted validate accuracy performance results promising when compared other conventional approaches models. model used create an automated floating detection system required timeframe. In both cases, able correctly achieved 98.38%. technique presented this can be crucial instrument for automatic garbage ocean thereby enhancing war against pollution.

Language: Английский

---

Using Fuzzy Logic to Analyse Weather Conditions

Electronics, Journal Year: 2024, Volume and Issue: 14(1), P. 85 - 85

Published: Dec. 28, 2024

Effective weather analysis is a very important scientific, social, and economic issue, because directly affects our lives has significant impact on various sectors, including agriculture, transport, energy, natural disaster management. Weather therefore the basis for operation of many decision-making support systems, especially in transport (air, sea), ensuring continuity supply chains industry or delivery food medicines, but also municipal economies tourism. Its role importance will grow with worsening climatic phenomena development Industry5.0 paradigm, which puts humans their environment at center attention. This article presents issues related to fuzzy sets systems model based them. The system was created using Matlab, Fuzzy Logic Designer application, focusing logic. With Designer, users can define sets, rules, carry out fuzzification defuzzification processes, thereby offering great possibilities data

Language: Английский

---

Machine learning model for Water Quality evaluation: Systematic Review

Published: Aug. 18, 2023

Evaluating water quality is essential to maintaining healthy ecosystems and providing safe drinking water. There has been a rise in enthusiasm for creating models evaluating thanks the development of machine learning methods. In order highlight most important results trends field evaluation, this systematic review seeks offer an overview advanced currently use. Relevant papers published during [detailed time period] were included after thorough search electronic resources. The studies chosen because their comprehensive coverage wide variety types characteristics, including pH, dissolved oxygen, turbidity, nutrient concentrations, across range sources (rivers, lakes, reservoirs, groundwater). This sheds light on algorithms used assessment. These from more traditional support vector machines (SVM) deep like convolutional neural networks (CNN) recurrent (RNN). Water metrics pollution may now be correctly predicted using these models. also delves into physicochemical meteorological data, topographical qualities, remote sensing data that go making work. encouraging progress toward better prediction part inclusion sources. Accuracy, precision, recall, correlation coefficients are only few performance evaluation criteria studied here. particular, it highlights shortages, concerns, model interpretability as bottlenecks creation rollout evaluation. Future research recommendations suggested findings review. interpretable models, consistent collecting sharing processes, improved stakeholder comprehension confidence outcomes all areas need attention. review's help further use techniques quality. They provide present state art, point out gaps, advice academics, policymakers, resource managers how best manage conserve

Language: Английский

---

Weather classification using meta-based random forest fusion of transfer learning models

International Journal of Advances in Intelligent Informatics, Journal Year: 2024, Volume and Issue: 10(2), P. 186 - 186

Published: May 31, 2024

Weather classification into multiple categories is an essential task for many applications, including farming, military, transport, airlines, navigation, agriculture, etc. A few pieces of research give attention to this field and the current state-of-art methods have limitations, low accuracy limited weather conditions. In study, a new meta-based fusion transfer deep learning model introduced. The study takes account all possible conditions utilizes technique improve performance. First, images are pre-processed data augmentation process performed. These fed five models (XceptionNet, VGG16, ResNet50V2, InceptionV3, DenseNet201). Then, random forest fusion, bagging score-level applied. Finally, individual evaluated. Experiments were conducted on WEAPD dataset which includes 11 categories. Results prove that best performance related ransom method with 96% accuracy. also compared methods, comparison proves robustness high especially fact achieves studies worked same dataset. RF promising methodology address problem. This outcome can be used by future ensemble methodologies.

Language: Английский

---

Continuous Monitoring of Water Levels for Industrial Boilers Using Single‐Stage Object Recognition YOLOv5

International Journal of Energy Research, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

This paper presents a measurement method that utilizes object recognition technology for continuous and quantitative real‐time monitoring of water levels in industrial boilers. Real‐time videos were monitored using small camera, the YOLO algorithm, single‐stage detector, was employed to use bounding boxes detected objects within video as variables, directly measuring length ratio each frame. The demonstrated high level accuracy water‐level measurement, with an average 99.02%, stable performance, fluctuation 0.13% measurements. Consequently, proposed proves feasible quantifying inspection systems even low‐resource environments. These results demonstrate new mechanism technology, without requiring text detection, showing potential improving efficiency complex boiler feasibility reliable control.

Language: Английский

---

Traffic forecasting using LSTM and SARIMA models: A comparative analysis

AIP conference proceedings, Journal Year: 2024, Volume and Issue: 3217, P. 020032 - 020032

Published: Jan. 1, 2024

Language: Английский

---