Investigating the Impact of Train / Test Split Ratio on the Performance of Pre-Trained Models with Custom Datasets

International Journal of Advanced Computer Science and Applications, Journal Year: 2024, Volume and Issue: 15(2)

Published: Jan. 1, 2024

The proper allocation of data between training and testing is a critical factor influencing the performance deep learning models, especially those built upon pre-trained architectures. Having suitable set size an important for classification model’s generalization performance. main goal this study to find appropriate three networks using different custom datasets. For aim, presented in paper explores effect varying train / test split ratio on popular namely MobileNetV2, ResNet50v2 VGG19, with focus image task. In work, balanced datasets never seen by models have been used, each containing 1000 images divided into two classes. ratios used are: 60-40, 70-30, 80-20 90-10. was metrics sensitivity, specificity overall accuracy evaluate classifiers under ratios. Experimental results show that, affected Moreover, more than 70% dataset task gives better

Language: Английский

---

A comparative assessment of machine learning algorithms with the Least Absolute Shrinkage and Selection Operator for breast cancer detection and prediction

Decision Analytics Journal, Journal Year: 2023, Volume and Issue: 7, P. 100245 - 100245

Published: May 6, 2023

Breast cancer is the most common life-threatening in women and one of leading causes death. Early diagnosis best defenses against spread breast cancer. Machine learning (ML) tools are now available for detection prediction. This study presents a comparative assessment machine models diagnosing based on various classification schemes. Our methodology well-organized data collection, preparation, transformation, exploratory analysis (including correlation matrix, histogram, distribution). All characteristics compared with results applying Least Absolute Shrinkage Selection Operator (LASSO) approach, which selects important attributes. Logistic Regression (LR), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGB), (GB), Random Forest (RF), Multilayer Perceptron (MLP), Support Vector (SVM) algorithms have been applied this study. We achieved maximum accuracy 90.68% by RF LASSO. Similarly, recall KNN was 98.80%, precision MLP 92.50%, F1 score 94.60%.

Language: Английский

---

Detection of Monkeypox Disease from Human Skin Images with a Hybrid Deep Learning Model

Diagnostics, Journal Year: 2023, Volume and Issue: 13(10), P. 1772 - 1772

Published: May 17, 2023

Monkeypox, a virus transmitted from animals to humans, is DNA with two distinct genetic lineages in central and eastern Africa. In addition zootonic transmission through direct contact the body fluids blood of infected animals, monkeypox can also be person skin lesions respiratory secretions an person. Various occur on individuals. This study has developed hybrid artificial intelligence system detect images. An open source image dataset was used for multi-class structure consisting chickenpox, measles, normal classes. The data distribution classes original unbalanced. augmentation preprocessing operations were applied overcome this imbalance. After these operations, CSPDarkNet, InceptionV4, MnasNet, MobileNetV3, RepVGG, SE-ResNet Xception, which are state-of-the-art deep learning models, detection. order improve classification results obtained unique model specific created by using highest-performing models long short-term memory (LSTM) together. proposed detection, test accuracy 87% Cohen's kappa score 0.8222.

Language: Английский

---

Monkeypox Diagnosis With Interpretable Deep Learning

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 81965 - 81980

Published: Jan. 1, 2023

As the world gradually recovers from impacts of COVID-19, recent global spread Monkeypox disease has raised concerns about another potential pandemic, highlighting urgency early detection and intervention to curb its transmission. Deep Learning (DL) based prediction presents a promising solution, offering affordable accessible diagnostic services. In this study, we harnessed Transfer (TL) techniques tweak assess performance an array six different DL models, encompassing VGG16, InceptionResNetV2, ResNet50, ResNet101, MobileNetV2, VGG19, Vision Transformer (ViT). Among diverse collection, it was modified versions VGG19 MobileNetV2 models that outshone others, boasting striking accuracy rates ranging impressive 93% astounding 99%. Our results echo findings research endeavours similarly showcased enhanced when developing armed with power TL. To add this, made use Local Interpretable Model Agnostic Explanations (LIME) lend sense transparency our model's predictions, identify crucial features correlating onset disease. These offer significant implications for prevention control efforts, particularly in remote resource-limited areas.

Language: Английский

---

CGO-ensemble: Chaos game optimization algorithm-based fusion of deep neural networks for accurate Mpox detection

Neural Networks, Journal Year: 2024, Volume and Issue: 173, P. 106183 - 106183

Published: Feb. 16, 2024

Language: Английский

---

Federated Deep Learning for Monkeypox Disease Detection on GAN-Augmented Dataset

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 32819 - 32829

Published: Jan. 1, 2024

After the coronavirus disease 2019 (COVID-19) outbreak, viral infection known as monkeypox gained significant attention, and World Health Organization (WHO) classified it a global public health emergency. Given similarities between other pox viruses, conventional classification methods encounter difficulties in accurately identifying disease. Furthermore, sharing sensitive medical data gives rise to concerns about security privacy. Integrating deep neural networks with federated learning (FL) presents promising avenue for addressing challenges of categorization. In light this, we propose an FL-based framework using models classify viruses securely. The proposed has three major components: (a) cycle-consistent generative adversarial network augment samples training; (b) learning-based such MobileNetV2, Vision Transformer (ViT), ResNet50 classification; (c) flower-federated environment security. experiments are performed publicly available datasets. experiments, ViT-B32 model yields impressive accuracy rate 97.90%, emphasizing robustness its potential secure accurate categorization

Language: Английский

---

Attention to Monkeypox: An Interpretable Monkeypox Detection Technique Using Attention Mechanism

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 51942 - 51965

Published: Jan. 1, 2024

In the wake of COVID-19, rising monkeypox cases pose a potential pandemic threat. While less severe than its increasing spread underscores urgency early detection and isolation to control disease. The main difficulty in diagnosing arises from prolonged diagnostic process symptoms that are similar those other skin diseases, making challenging. To address this, deployment deep learning models on edge devices presents viable solution for rapid accurate monkeypox. However, resource constraints require use lightweight models. limitation these often involves trade-off with accuracy, which is unacceptable context medical diagnostics. Therefore, development optimized both resource-efficient computing highly becomes imperative. this end, an attention-based MobileNetV2 model detection, capitalizing inherent design effective devices, proposed. This model, enhanced spatial channel attention mechanisms, tailored early-stage diagnosis better accuracy. We significantly improved Monkeypox Skin Images Dataset (MSID) by incorporating broader range classes thereby substantially enriching diversifying training dataset. helps distinguish particularly stages or when detailed examination unavailable. ensure transparency interpretability, we incorporated Gradient-weighted Class Activation Mapping (Grad-CAM) Local Interpretable Model-Agnostic Explanations (LIME) provide clear insights into model's reasoning. Finally, comprehensively assess performance our employed evaluation metrics, including Cohen's Kappa, Matthews Correlation Coefficient, Youden's J Index, alongside traditional measures like F1-score, precision, recall, sensitivity, specificity. demonstrated impressive results, outperforming baseline achieving 92.28% accuracy extended MSID dataset, 98.19% original 93.33% Lesion (MSLD)

Language: Английский

---

AI-Based Approaches for the Diagnosis of Mpox: Challenges and Future Prospects

Archives of Computational Methods in Engineering, Journal Year: 2024, Volume and Issue: 31(6), P. 3585 - 3617

Published: March 26, 2024

Language: Английский

---

Deep and Transfer Learning Approaches for Automated Early Detection of Monkeypox (Mpox) Alongside Other Similar Skin Lesions and Their Classification

ACS Omega, Journal Year: 2023, Volume and Issue: 8(35), P. 31747 - 31757

Published: Aug. 23, 2023

The world faces multiple public health emergencies simultaneously, such as COVID-19 and Monkeypox (mpox). mpox, from being a neglected disease, has emerged global threat that spread to more than 100 nonendemic countries, even been spreading for 3 years now. general mpox symptoms are similar chickenpox measles, thus leading possible misdiagnosis. This study aimed at facilitating rapid high-brevity diagnosis. Reportedly, circulates among particular groups, sexually promiscuous gay bisexuals. Hence, selectively vaccinating, isolating, treating them seems difficult due the associated social stigma. Deep learning (DL) great promise in image-based diagnosis could help error-free bulk novelty proposed, system adopted, methods approaches discussed article. present work proposes use of DL models automated early performances proposed algorithms were evaluated using data set available domain. adopted was meant both training testing, details which elaborated. CNN, VGG19, ResNet 50, Inception v3, Autoencoder compared. It concluded v3 detection skin lesions, returned best (96.56%) classification accuracy.

Language: Английский

---

Medicare Fraud Detection Using Graph Analysis: A Comparative Study of Machine Learning and Graph Neural Networks

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 88278 - 88294

Published: Jan. 1, 2023

Insurance companies have focused on medicare fraud detection to reduce financial losses and reputational harm because causes tens of billions dollars in damage annually. This study demonstrates that can be significantly enhanced by introducing graph analysis with considering the relationships among medical providers, beneficiaries, physicians. We use open-source tabular datasets containing beneficiary information, inpatient claims, outpatient indications about potential fraudulent providers. then aggregated them into a single dataset converting structure. Furthermore, we developed models using two approaches reflect i.e., neural network (GNN) traditional machine learning centrality measures. Therefore, model features showed improved precision 4 percent point (%p), recall 24 %p, F1-score 14 %p compared best GNN model. The improvement this extent could result substantial cost savings 3.1 billion euros 5 United States Europe, respectively, benefiting governmental institutions insurance involved healthcare operations. required time was approximately 250-300 times more than machine-learning outcome suggests successful efficient achieved if measures are used capture physicians, beneficiaries.

Language: Английский

---