A Deep Learning Ensemble With Data Resampling for Credit Card Fraud Detection

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 30628 - 30638

Published: Jan. 1, 2023

Credit cards play an essential role in today's digital economy, and their usage has recently grown tremendously, accompanied by a corresponding increase credit card fraud. Machine learning (ML) algorithms have been utilized for fraud detection. However, the dynamic shopping patterns of holders class imbalance problem made it difficult ML classifiers to achieve optimal performance. In order solve this problem, paper proposes robust deep-learning approach that consists long short-term memory (LSTM) gated recurrent unit (GRU) neural networks as base learners stacking ensemble framework, with multilayer perceptron (MLP) meta-learner. Meanwhile, hybrid synthetic minority oversampling technique edited nearest neighbor (SMOTE-ENN) method is employed balance distribution dataset. The experimental results showed combining proposed deep SMOTE-ENN achieved sensitivity specificity 1.000 0.997, respectively, which superior other widely used methods literature.

Language: Английский

---

Increasing transparency in machine learning through bootstrap simulation and shapely additive explanations

PLoS ONE, Journal Year: 2023, Volume and Issue: 18(2), P. e0281922 - e0281922

Published: Feb. 23, 2023

Machine learning methods are widely used within the medical field. However, reliability and efficacy of these models is difficult to assess, making it for researchers identify which machine-learning model apply their dataset. We assessed whether variance calculations metrics (e.g., AUROC, Sensitivity, Specificity) through bootstrap simulation SHapely Additive exPlanations (SHAP) could increase transparency improve selection. Data from England National Health Services Heart Disease Prediction Cohort was used. After comparison XGBoost, Random Forest, Artificial Neural Network, Adaptive Boosting, XGBoost as choice in this study. Boost-strap (N = 10,000) empirically derive distribution covariate Gain statistics. provide explanations output evaluate accuracy metrics. For modeling method, we observed (through 10,000 completed simulations) that AUROC ranged 0.771 0.947, a difference 0.176, balanced 0.688 0.894, 0.205 difference, sensitivity 0.632 0.939, 0.307 specificity 0.595 0.944, 0.394 difference. Among simulations completed, gain Angina 0.225 0.456, 0.231, Cholesterol 0.148 0.326, 0.178, maximum heart rate (MaxHR) 0.081 0.200, range 0.119, Age 0.059 0.157, 0.098. Use variability explanatory algorithms observe if covariates match literature necessary increased transparency, reliability, utility machine methods. These statistics, combined with statistics can help best given

Language: Английский

---

Supervised machine learning in drug discovery and development: Algorithms, applications, challenges, and prospects

Machine Learning with Applications, Journal Year: 2024, Volume and Issue: 17, P. 100576 - 100576

Published: July 24, 2024

Language: Английский

---

Influence of Optimal Hyperparameters on the Performance of Machine Learning Algorithms for Predicting Heart Disease

Processes, Journal Year: 2023, Volume and Issue: 11(3), P. 734 - 734

Published: March 1, 2023

One of the most difficult challenges in medicine is predicting heart disease at an early stage. In this study, six machine learning (ML) algorithms, viz., logistic regression, K-nearest neighbor, support vector machine, decision tree, random forest classifier, and extreme gradient boosting, were used to analyze two datasets. dataset was UCI Kaggle Cleveland other comprehensive Cleveland, Hungary, Switzerland, Long Beach V. The performance results techniques obtained. with tuned hyperparameters achieved highest testing accuracy 87.91% for dataset-I boosting classifier 99.03% dataset-II. novelty work use grid search cross-validation enhance form training testing. ideal parameters identified through experimental results. Comparative studies also carried out existing focusing on prediction disease, where approach significantly outperformed their

Language: Английский

---

A Machine Learning Method with Hybrid Feature Selection for Improved Credit Card Fraud Detection

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(12), P. 7254 - 7254

Published: June 18, 2023

With the rapid developments in electronic commerce and digital payment technologies, credit card transactions have increased significantly. Machine learning (ML) has been vital analyzing customer data to detect prevent fraud. However, presence of redundant irrelevant features most real-world degrades performance ML classifiers. This study proposes a hybrid feature-selection technique consisting filter wrapper steps ensure that only relevant are used for machine learning. The proposed method uses information gain (IG) rank features, top-ranked fed genetic algorithm (GA) wrapper, which extreme (ELM) as algorithm. Meanwhile, GA is optimized imbalanced classification using geometric mean (G-mean) fitness function instead conventional accuracy metric. approach achieved sensitivity specificity 0.997 0.994, respectively, outperforming other baseline techniques methods recent literature.

Language: Английский

---

A deep learning approach for Maize Lethal Necrosis and Maize Streak Virus disease detection

Machine Learning with Applications, Journal Year: 2024, Volume and Issue: 16, P. 100556 - 100556

Published: May 7, 2024

Maize is an important crop cultivated in Sub-Saharan Africa, essential for food security. However, its cultivation faces significant challenges due to debilitating diseases such as Lethal Necrosis (MLN) and Streak Virus (MSV), which can lead severe yield losses. Traditional plant disease diagnosis methods are often time-consuming prone errors, necessitating more efficient approaches. This study explores the application of deep learning, specifically Convolutional Neural Networks (CNNs), automatic detection classification maize diseases. We investigate six architectures: Basic CNN, EfficientNet V2 B0 B1, LeNet-5, VGG-16, ResNet50, using a dataset 15344 images comprising MSV, MLN, healthy leaves. Additionally, performed hyperparameter tuning improve performance models Gradient-weighted Class Activation Mapping (Grad-CAM) model interpretability. Our results show that demonstrated accuracy 99.99% distinguishing between disease-infected plants. The this contribute advancement AI applications agriculture, particularly diagnosing within Africa.

Language: Английский

---

An Improved Ensemble Method for Predicting Hyperchloremia in Adults With Diabetic Ketoacidosis

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 9536 - 9549

Published: Jan. 1, 2024

Diabetic ketoacidosis (DKA) is a serious complication that affects millions of individuals globally and presents significant health complications. Hyperchloremia, an electrolyte imbalance characterized by high levels chloride in the blood, may result gastrointestinal problems, kidney damage, even death, especially DKA patients. Early detection treatment hyperchloremia are utmost importance management DKA. This study explores potential bootstrap aggregating ensemble with random subspaces machine learning approach to predict occurrence hyperchloremia, providing basis for early intervention improved patient outcomes. We tested our retrospective MIMIC-III database containing 1177 patients compared it previous studies area under curve (AUC) 100%. Our showed performance outperforming other methods. The combination this enhance timely cases, ultimately leading outcomes more effective DKA-associated work aims contribute development decision support tools healthcare professionals, assisting them making informed decisions patients, focus on preventing managing hyperchloremia.

Language: Английский

---

An Improved Framework for Detecting Thyroid Disease Using Filter-Based Feature Selection and Stacking Ensemble

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 89098 - 89112

Published: Jan. 1, 2024

Language: Английский

---

Intelligent Internet of Medical Things for Depression: Current Advancements, Challenges, and Trends

International Journal of Intelligent Systems, Journal Year: 2025, Volume and Issue: 2025(1)

Published: Jan. 1, 2025

We investigated the fusion of Intelligent Internet Medical Things (IIoMT) with depression management, aiming to autonomously identify, monitor, and offer accurate advice without direct professional intervention. Addressing pivotal questions regarding IIoMT’s role in identification, its correlation stress anxiety, impact machine learning (ML) deep (DL) on depressive disorders, challenges potential prospects integrating management IIoMT, this research offers significant contributions. It integrates artificial intelligence (AI) (IoT) paradigms expand studies, highlighting data science modeling’s practical application for intelligent service delivery real‐world settings, emphasizing benefits within IoT. Furthermore, it outlines an IIoMT architecture gathering, analyzing, preempting employing advanced analytics enhance intelligence. The study also identifies current challenges, future trajectories, solutions domain, contributing scientific understanding management. evaluates 168 closely related articles from various databases, including Web Science (WoS) Google Scholar, after rejection repeated books. shows that there is 48% growth articles, mainly focusing symptoms, detection, classification. Similarly, most being conducted United States America, trend increasing other countries around globe. These results suggest essence automated monitoring, suggestions handling depression.

Language: Английский

---

Vaccine Hesitancy Hotspots in Africa: An Insight From Geotagged Twitter Posts

IEEE Transactions on Computational Social Systems, Journal Year: 2023, Volume and Issue: 11(1), P. 1325 - 1338

Published: Jan. 19, 2023

Many social media users express concerns about vaccines and their side effects on Twitter. These lead to a compromise of confidence which brings vaccine hesitancy. In Africa, hesitancy is major challenge faced by health policymakers in the fight against COVID-19. Given that most tweets are geotagged, clustering them according sentiments could help identify locations may likely experience for policy planning. this study, we collected 70 000 geotagged vaccine-related nine African countries, from December 2020 February 2022. The were classified into three sentiment classes—positive, negative, neutral. quality classification outputs was achieved using Naíve Bayes (NB), logistic regression (LR), support vector machines (SVMs), decision tree (DT), K-nearest neighbor (KNN) machine learning classifiers. LR highest accuracy 71% with an average area under curve 85%. point-based location technique used calculate hotspots based tweets. Locations green, red, gray backgrounds map signify hotspot positive, neutral sentiments. outcome research shows discussions can be analyzed during disease outbreak, inform planning management Africa.

Language: Английский

---