Mobility-aware computational offloading in mobile edge networks: a survey

Cluster Computing, Journal Year: 2021, Volume and Issue: 24(4), P. 2735 - 2756

Published: April 9, 2021

Language: Английский

---

Performance Analysis of Machine Learning Centered Workload Prediction Models for Cloud

IEEE Transactions on Parallel and Distributed Systems, Journal Year: 2023, Volume and Issue: 34(4), P. 1313 - 1330

Published: Jan. 30, 2023

The precise estimation of resource usage is a complex and challenging issue due to the high variability dimensionality heterogeneous service types dynamic workloads. Over last few years, prediction traffic has received ample attention from research community. Many machine learning-based workload forecasting models have been developed by exploiting their computational power learning capabilities. This paper presents first systematic survey cum performance analysis-based comparative study diversified learning-driven cloud models. discussion initiates with significance predictive management followed schematic description, operational design, motivation, challenges concerning these Classification taxonomy different approaches into five distinct categories are presented focusing on theoretical concepts mathematical functioning existing state-of-the-art methods. most prominent belonging class thoroughly surveyed compared. All classified implemented common platform for investigation comparison using three benchmark traces via experimental analysis. essential key indicators evaluated concluded discussing trade-offs notable remarks.

Language: Английский

---

MAG-D: A multivariate attention network based approach for cloud workload forecasting

Future Generation Computer Systems, Journal Year: 2023, Volume and Issue: 142, P. 376 - 392

Published: Jan. 10, 2023

Language: Английский

---

Predictive intelligence using ANFIS‐induced OWAWA for complex stock market prediction

International Journal of Intelligent Systems, Journal Year: 2021, Volume and Issue: 37(8), P. 4586 - 4611

Published: Nov. 8, 2021

Traditional time series prediction methods are unable to handle the complex nonlinear relationship of a large data set. Most existing techniques manage multiple dimensions set, due which computational complexity escalates with increasing size Many machine learning (ML) known unknown predictions. This paper presents new forecasting method in neural network structure based on induced ordered weighted average (IOWA) (WA) and fuzzy series. The proposed model is more efficient than handling other traditional methods. can accommodate IOWA operator, average, relevance degree each concept particular problem for prediction. contribution this twofold. First, it contributes theory by proposing IOWAWA layer second application approach predict stock market data. robustness tested using Australian Securities Exchange (ASX) considering case study housing property sector. We further compare accuracy sixteen experimental results demonstrate that outperforms

Language: Английский

---

Multivariate workload and resource prediction in cloud computing using CNN and GRU by attention mechanism

The Journal of Supercomputing, Journal Year: 2022, Volume and Issue: 79(3), P. 3437 - 3470

Published: Sept. 4, 2022

Language: Английский

---

A hybrid deep neural network approach to estimate reference evapotranspiration using limited climate data

Neural Computing and Applications, Journal Year: 2021, Volume and Issue: 34(5), P. 4013 - 4032

Published: Nov. 12, 2021

Language: Английский

---

Accurate Prediction of Workloads and Resources With Multi-Head Attention and Hybrid LSTM for Cloud Data Centers

IEEE Transactions on Sustainable Computing, Journal Year: 2023, Volume and Issue: 8(3), P. 375 - 384

Published: March 20, 2023

Currently, cloud computing service providers face big challenges in predicting large-scale workload and resource usage time series. Due to the difficulty capturing nonlinear features, traditional forecasting methods usually fail achieve high prediction performance for sequences. Besides, there is much noise original series of resources workloads. If these are not de-noised by smoothing algorithms, results can meet providers' requirements. To do so, this work proposes a hybrid model named VAMBiG that integrates V ariational mode decomposition, an xmlns:xlink="http://www.w3.org/1999/xlink">A daptive Savitzky-Golay (SG) filter, xmlns:xlink="http://www.w3.org/1999/xlink">M ulti-head attention mechanism, xmlns:xlink="http://www.w3.org/1999/xlink">Bi directional xmlns:xlink="http://www.w3.org/1999/xlink">G rid versions Long Short Term Memory (LSTM) networks. adopts signal decomposition method variational decompose complex non-linear into low-frequency intrinsic functions. Then, it adaptive SG filter as data pre-processing tool eliminate extreme points such Afterwards, bidirectional grid LSTM networks capture features dimension ones, respectively. Finally, multi-head mechanism explore importance different dimensions. aims predict workloads highly variable traces clouds. Extensive experimental demonstrate achieves higher-accuracy than several advanced approaches with datasets from Google Alibaba cluster traces.

Language: Английский

---

Symmetry-Aware Multi-Dimensional Attention Spiking Neural Network with Optimization Techniques for Accurate Workload and Resource Time Series Prediction in Cloud Computing Systems

Symmetry, Journal Year: 2025, Volume and Issue: 17(3), P. 383 - 383

Published: March 3, 2025

Cloud computing offers scalable and adaptable resources on demand, has emerged as an essential technology for contemporary enterprises. Nevertheless, it is still challenging work to efficiently handle cloud because of dynamic changes in load requirement. Existing forecasting approaches are unable the intricate temporal symmetries nonlinear patterns workload data, leading degradation prediction accuracy. In this manuscript, a Symmetry-Aware Multi-Dimensional Attention Spiking Neural Network with Optimization Techniques Accurate Workload Resource Time Series Prediction Computing Systems (MASNN-WL-RTSP-CS) proposed. Here, input data from Google cluster trace dataset were preprocessed using Multi Window Savitzky–Golay Filter (MWSGF) remove noise while preserving important maintaining structural symmetry time series trends. Then, (MASNN) effectively models symmetric fluctuations predict resource series. To enhance accuracy, Secretary Bird Algorithm (SBOA) was utilized optimize MASNN parameters, ensuring accurate predictions. Experimental results show that MASNN-WL-RTSP-CS method achieves 35.66%, 32.73%, 31.43% lower Root Mean Squared Logarithmic Error (RMSLE), 25.49%, 32.77%, 28.93% Square (MSE), 24.54%, 23.65%, 23.62% Absolute (MAE) compared other approaches, like ICNN-WL-RP-CS, PA-ENN-WLP-CS, DCRNN-RUP-RP-CCE, respectively. These advances emphasize utility achieving more forecasts, thereby facilitating effective management.

Language: Английский

---

A temporal evolution and fine-grained information aggregation model for citation count prediction

Scientometrics, Journal Year: 2025, Volume and Issue: unknown

Published: April 9, 2025

Language: Английский

---

Deep learning approaches for time series prediction in climate resilience applications

Frontiers in Environmental Science, Journal Year: 2025, Volume and Issue: 13

Published: April 28, 2025

Introduction Time series prediction is a fundamental task in climate resilience, where accurate forecasting of variables critical for proactive planning and adaptation. Traditional methods often struggle with the nonlinearity, high variability, multi-scale dependencies inherent data, limiting their applicability dynamic diverse environments. Methods In this work, we propose novel framework that combines Resilience Optimization Network (ResOptNet) Equity-Driven Climate Adaptation Strategy (ED-CAS) to address these challenges. ResOptNet employs hybrid predictive modeling multi-objective optimization identify tailored interventions risk mitigation, dynamically adapting real-time data through feedback-driven loop. ED-CAS complements by embedding equity considerations into resource allocation, ensuring resilience-building efforts prioritize vulnerable populations regions. Results Experimental evaluations on datasets demonstrate our approach significantly improves accuracy, resilience indices, equitable distribution compared traditional models. Discussion By integrating analytics equity-driven strategies, provides actionable insights adaptation, advancing development scalable socially just solutions.

Language: Английский

---