Optimized EWT-Seq2Seq-LSTM with Attention Mechanism to Insulators Fault Prediction

Sensors, Journal Year: 2023, Volume and Issue: 23(6), P. 3202 - 3202

Published: March 17, 2023

Insulators installed outdoors are vulnerable to the accumulation of contaminants on their surface, which raise conductivity and increase leakage current until a flashover occurs. To improve reliability electrical power system, it is possible evaluate development fault in relation thus predict whether shutdown may occur. This paper proposes use empirical wavelet transform (EWT) reduce influence non-representative variations combines attention mechanism with long short-term memory (LSTM) recurrent network for prediction. The Optuna framework has been applied hyperparameter optimization, resulting method called optimized EWT-Seq2Seq-LSTM attention. proposed model had 10.17% lower mean square error (MSE) than standard LSTM 5.36% MSE without showing that optimization promising strategy.

Language: Английский

---

KWHO-CNN: A Hybrid Metaheuristic Algorithm Based Optimzed Attention-Driven CNN for Automatic Clinical Depression Recognition

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2024, Volume and Issue: 10(3)

Published: Sept. 27, 2024

Depression is a widespread mental disorder with inconsistent symptoms that make diagnosis challenging in clinical practice and research. Nevertheless, the poor identification may be partially explained by fact present approaches ignore patients' vocal tract modifications favour of merely considering speech perception aspects. This study proposes novel framework, KWHO-CNN, integrating hybrid metaheuristic algorithm Attention-Driven Convolutional Neural Networks (CNNs), to enhance depression detection using data. It addresses challenges like variability patterns small sample sizes optimizing feature selection classification. Initial pre-processing involves noise reduction, data normalization, segmentation, followed extraction, primarily utilizing Mel-frequency cepstral coefficients (MFCCs). The Krill Wolf Hybrid Optimization (KWHO) Algorithm optimizes these features, overcoming issues over-fitting enhancing model performance. CNN architecture further refines classification, leveraging dense computations architectural homogeneity. suggested outperforms diagnosis, over 90% accuracy, precision, recall, F1 score, demonstrating its potential greatly impact health

Language: Английский

---

Recognition model for major depressive disorder in Arabic user-generated content

Beni-Suef University Journal of Basic and Applied Sciences, Journal Year: 2025, Volume and Issue: 14(1)

Published: Jan. 24, 2025

Abstract Background One of the psychological problems that have become very prevalent in modern world is depression, where mental health disorders common. Depression, as reported by WHO, second-largest factor worldwide burden illnesses. As these issues grow, social media has a tremendous platform for people to express themselves. A user’s behavior may therefore disclose lot about their emotional state and health. This research offers novel framework depression detection from Arabic textual data utilizing deep learning (DL), natural language processing (NLP), machine (ML), BERT transformers techniques light disease’s high prevalence. To do this, dataset tweets was used, which collected 3 sources, we mention later. The constructed two variants, one with binary classification other multi-classification. Results In classifications, used ML such “support vector (SVM), random forest (RF), logistic regression (LR), Gaussian naive Bayes (GNB),” “ARABERT.” comparison transformers, ARABERT accuracy 93.03 percent rate. multi-classification, DL “long short-term memory (LSTM),” “Multilingual BERT.” multilingual multi-classification an 97.8%. Conclusion Through user-generated content, can detect depressed using artificial intelligence technology fast manner instead medical technology.

Language: Английский

---

Cost-effective time-efficient subnational-level surveillance using Twitter: Kingdom of Saudi Arabia case study

Deleted Journal, Journal Year: 2025, Volume and Issue: 7(1)

Published: Jan. 8, 2025

An effective Twitter-based surveillance system should provide insights at national and subnational levels. The literature identifies two methodologies for geolocating tweets: using only geotagged tweets or retrieving all relevant tweets, then filtering out those not belonging to the target geographical region. first methodology is accurate, cost-effective, time-efficient but has limited coverage. second offers better coverage less particularly informal Arabic text, neither cost-effective nor due Twitter's new policies. There a gap in an solution with reasonable To fill this gap, we propose that uses underutilized feature Twitter backend geolocate during data collection. This retrieves both geolocated ensuring accuracy It also as are retrieved. Applying Saudi Arabia COVID-19, generated dataset, KSAGeoCOV, 4.25 times more than geotagged-only dataset. successfully predicted COVID-19 outbreaks June 2021 January 2022. Pearson correlation coefficient between WHO weekly reported cases returned 1-week lag, $$r = 0.733;\,p < 0.001$$ 0.814;\,p when including English indicating very strong level. At level, top-populated provinces show correlations ( 0.64$$ $$0.74;\,p 0.003$$ ).

Language: Английский

---

Attention-Enabled Ensemble Deep Learning Models and Their Validation for Depression Detection: A Domain Adoption Paradigm

Diagnostics, Journal Year: 2023, Volume and Issue: 13(12), P. 2092 - 2092

Published: June 16, 2023

Depression is increasingly prevalent, leading to higher suicide risk. detection and sentimental analysis of text inputs in cross-domain frameworks are challenging. Solo deep learning (SDL) ensemble (EDL) models not robust enough. Recently, attention mechanisms have been introduced SDL. We hypothesize that attention-enabled EDL (aeEDL) architectures superior compared attention-not-enabled SDL (aneSDL) or aeSDL models. designed EDL-based with blocks build eleven kinds model five on four domain-specific datasets. scientifically validated our by comparing "seen" "unseen" paradigms (SUP). benchmarked results against the SemEval (2016) dataset established reliability tests. The mean increase accuracy for over their corresponding components was 4.49%. Regarding effect block, (AUC) aneSDL 2.58% (1.73%), aeEDL aneEDL 2.76% (2.80%). When vs. non-attention attention, greater than 4.82% (3.71%), 5.06% (4.81%). For benchmarking (SemEval), best-performing (ALBERT+BERT-BiLSTM) best (BERT-BiLSTM) 3.86%. Our scientific validation design showed a difference only 2.7% SUP, thereby meeting regulatory constraints. all hypotheses further demonstrated very effective generalized method detecting symptoms depression settings.

Language: Английский

---

Depressive Disorder Recognition Based on Frontal EEG Signals and Deep Learning

Sensors, Journal Year: 2023, Volume and Issue: 23(20), P. 8639 - 8639

Published: Oct. 23, 2023

Depressive disorder (DD) has become one of the most common mental diseases, seriously endangering both affected person’s psychological and physical health. Nowadays, a DD diagnosis mainly relies on experience clinical psychiatrists subjective scales, lacking objective, accurate, practical, automatic technologies. Recently, electroencephalogram (EEG) signals have been widely applied for diagnosis, but with high-density EEG, which can severely limit efficiency EEG data acquisition reduce practicability diagnostic techniques. The current study attempts to achieve accurate practical diagnoses based combining frontal six-channel deep learning models. To this end, 10 min resting-state were collected from 41 patients 34 healthy controls (HCs). Two models, multi-resolution convolutional neural network (MRCNN) combined long short-term memory (LSTM) (named MRCNN-LSTM) MRCNN residual squeeze excitation (RSE) MRCNN-RSE), proposed recognition. results showed that higher frequency band obtained better classification performance diagnosis. MRCNN-RSE model achieved highest accuracy 98.48 ± 0.22% 8–30 Hz signals. These findings indicated analytical framework provide an strategy as well essential theoretical technical support treatment efficacy evaluation DD.

Language: Английский

---

RNN-CNN Based Hybrid Deep Learning Model for Mental Healthcare

Algorithms for intelligent systems, Journal Year: 2025, Volume and Issue: unknown, P. 411 - 424

Published: Jan. 1, 2025

Language: Английский

---

Sentiment analysis applications using deep learning advancements in social networks: A systematic review

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 129862 - 129862

Published: March 1, 2025

Language: Английский

---

Lightweight Multi‐Stage Holistic Attention‐Based Network for Image Super‐Resolution

IET Image Processing, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT High‐resolution images are crucial for many applications, but factors such as environmental conditions can reduce image quality. Super‐resolution (SR) techniques address this by generating high‐resolution from low‐resolution inputs. While deep learning SR models have made significant progress, they be computationally expensive and struggle with differentiating between various scales. Lightweight methods, suitable resource‐constrained devices, often compromise This study introduces a multi‐stage holistic attention‐based network, using Gaussian Laplacian pyramids to decompose apply attention modules at each level. approach reduces parameters computational costs while maintaining quality, achieving PSNR score of 28 SSIM 0.91 only 29,000 parameters. The model demonstrates the potential efficient high‐quality reconstruction. Future work will focus on improving quality minimizing exploring other advanced techniques. code available upon request

Language: Английский

---

A systematic approach to detect the emotions from the text using novel bidirectional encoder representation from transformers algorithm and compare the accuracy with long short-term memory algorithm

AIP conference proceedings, Journal Year: 2025, Volume and Issue: 3270, P. 020122 - 020122

Published: Jan. 1, 2025

Language: Английский

---