Predicting Chronic Liver Disease Using Boosting Technique

Published: Dec. 29, 2023

Liver disease has become a major health crisis globally. Machine learning methodologies are increasingly being applied to predict and diagnose various diseases. This paper uses five boosting algorithms (XGBoost, CatBoost, LightGBM, AdaBoost, gradient boosting) liver disease. Several preprocessing procedures utilised enhance the prediction performance, in addition appropriate tuning of hyperparameters selection features. The model's performance is assessed using metrics, including accuracy, precision, recall, fl-score, misclassification rate, AVC-ROC, runtime. Among methods evaluated, emerged as best performer, attaining highest scores nearly all metrics. It achieved an AVC-ROC 86%, accuracy 87.43%, precision recall 88.5%.

Language: Английский

---

Classification of skin cancer using deep batch-normalized elu alexnet with fractional sparrow ladybug optimization

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(14), P. 42319 - 42347

Published: Oct. 16, 2023

Language: Английский

---

Employing a Hybrid Convolutional Neural Network and Extreme Learning Machine for Precision Liver Disease Forecasting

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

Published: Jan. 1, 2024

This paper discusses the critical relevance of precise forecasting in liver disease, as well need for early identification and categorization immediate action personalized treatment strategies. The describes a unique strategy improving disease classification using ultrasound image processing. recommended technique combines properties Extreme Learning Machine (ELM), Convolutional Neural Network (CNN), along Grey Wolf Optimisation (GWO) to form an integrated model known CNN-ELM-GWO. data is provided by Pakistan's Multan Institute Nuclear Medicine Radiotherapy, it then pre-processed utilizing bilateral optimal wavelet filtering techniques increase dataset's quality. To properly extract significant visual information, feature extraction employs deep CNN architecture six convolutional layers, batch normalization, max-pooling. ELM serves classifier, whereas extractor. GWO algorithm, based on grey wolf searching strategies, refines hyperparameters two stages, progressively boosting system's accuracy. When implemented Python, CNN-ELM-GWO exceeds traditional machine learning algorithms (MLP, RF, KNN, NB) terms accuracy, precision, recall, F1-score metrics. proposed achieves impressive 99.7% revealing its potential significantly enhance employing images. outperforms conventional approaches substantial margin 27.5%, showing revolutionize medical imaging prospects.

Language: Английский

---

Optimizing Insurance Fraud Claim Detection through Machine Learning: A Comprehensive Approach for Improved Fraud Detection

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: March 19, 2024

Abstract Insurance fraud is a growing concern, prompting proactive measures through advanced machine learning techniques. This research focuses on constructing predictive model for distinguishing genuine and fraudulent auto insurance claims. The dataset, comprising 1,000 instances 40 attributes, covers customer demographics, policy details, incidents, financial data. Early detection crucial loss mitigation maintaining system integrity. study employs data preprocessing to handle missing values features XGBoost importance, variance thresholding, correlation analysis enhanced interpretability. integrates nine algorithms, with hard-voting ensemble of Logistic Regression demonstrating competitive accuracy, reaching 83.0%. Results highlight Linear Discriminant Analysis as the leading classifier, achieving 84% accuracy. approach achieves 83.0% accuracy notable precision 91%, showcasing strength combining diverse models. emphasizes significance preprocessing, feature selection, optimization. refined minimal Brier 0.00054, indicating discrepancies in predicted probabilities actual outcomes binary classification. Exploration principal component (PCA) multiple linear regression reveals trade-off between simplicity performance. Retaining 32 components preserves 95% variance, balance at 0.7967, while keeping 35 reaches highest value 0.9991, dimensionality reduction's potential capture nearly all variance.

Language: Английский

---

Discriminating insulin resistance in middle-aged nondiabetic women using machine learning approaches

AIMS Public Health, Journal Year: 2024, Volume and Issue: 11(2), P. 667 - 687

Published: Jan. 1, 2024

<abstract><sec> <title>Objective</title> <p>We employed machine learning algorithms to discriminate insulin resistance (IR) in middle-aged nondiabetic women.</p> </sec><sec> <title>Methods</title> <p>The data was from the National Health and Nutrition Examination Survey (2007–2018). The study subjects were 2084 women aged 45–64. analysis included 48 predictors. We randomly divided into training (n = 1667) testing 417) datasets. Four techniques IR: extreme gradient boosting (XGBoosting), random forest (RF), (GBM), decision tree (DT). area under curve (AUC) of receiver operating characteristic (ROC), accuracy, sensitivity, specificity, positive predictive value, negative F1 score compared as performance metrics select optimal technique.</p> <title>Results</title> XGBoosting algorithm achieved a relatively high AUC 0.93 dataset 0.86 IR using predictors followed by RF, GBM, DT models. After selecting top five build models, XGBoost with 0.90 (training dataset) (testing remained prediction model. SHapley Additive exPlanations (SHAP) values revealed associations between IR, namely BMI (strongly impact on IR), fasting glucose positive), HDL-C (medium negative), triglycerides glycohemoglobin positive). threshold for identifying 29 kg/m², 100 mg/dL, 54.5 89 5.6% BMI, glucose, HDL-C, triglycerides, glycohemoglobin, respectively.</p> <title>Conclusion</title> demonstrated superior discriminating women, predictors.</p> </sec></abstract>

Language: Английский

---