Improve Model Accuracy Research Articles

Detection of Elementary White Mucosal Lesions by an AI System: A Pilot Study

Aim: Accurately identifying primary lesions in oral medicine, particularly elementary white lesions, is a significant challenge, especially for trainee dentists. This study aimed to develop and evaluate a deep learning (DL) model for the detection and classification of elementary white mucosal lesions (EWMLs) using clinical images. Materials and Methods: A dataset was created by collecting photographs of various oral lesions, including oral leukoplakia, OLP plaque-like and reticular forms, OLL, oral candidiasis, and hyperkeratotic lesions from the Unit of Oral Medicine. The SentiSight.AI (Neurotechnology Co.®, Vilnius, Lithuania) AI platform was used for image labeling and model training. The dataset comprised 221 photos, divided into training (n = 179) and validation (n = 42) sets. Results: The model achieved an overall precision of 77.2%, sensitivity of 76.0%, F1 score of 74.4%, and mAP of 82.3%. Specific classes, such as condyloma and papilloma, demonstrated high performance, while others like leucoplakia showed room for improvement. Conclusions: The DL model showed promising results in detecting and classifying EWMLs, with significant potential for educational tools and clinical applications. Expanding the dataset and incorporating diverse image sources are essential for improving model accuracy and generalizability.

Urban drainage efficiency evaluation and flood simulation using integrated SWMM and terrain structural analysis.

A method for evaluating urban drainage system efficiency and simulating pluvial flooding is presented, incorporating rainwater inlet limitations through an integrated SWMM and terrain structural analysis (the corrected model). The corrected model is calibrated using data from two flood events and compared to an overflow model (non-corrected, without considering inlet restrictions) under the same conditions to assess its performance. The results show that the relative error of the flood peak in simulations ranges from 0.27 % to 2.80 %, while the Nash efficiency coefficient ranges from 0.3184 to 0.9504, indicating reasonable accuracy and significant comparability to the overflow model. Drainage efficiency evaluation reveals significant variation among drainage units, decreasing with increasing rainfall until surface depressions fill, after which surface drainage improves notably. This highlights the critical role of surface drainage in overall system efficiency and flood dynamics. Furthermore, two types of urban storm-induced floods are identified: infiltration excess and saturation excess floods. Modeling all floods as saturation excess may lead to substantial misestimations, stressing the need for accurate modeling that accounts for drainage unit characteristics. Understanding urban flood mechanisms is essential for improving modeling accuracy and optimizing drainage system design and flood management strategies.

Enhancing Cover Management Factor Classification Through Imbalanced Data Resolution

This study addresses the persistent challenge of class imbalance in land use and land cover (LULC) classification within the Shihmen Reservoir watershed in Taiwan, where LULC is used to map the Cover Management factor (C-factor). The dominance of forests in the LULC categories leads to an imbalanced dataset, resulting in poor prediction performance for minority classes when using machine learning techniques. To overcome this limitation, we applied the Synthetic Minority Over-sampling Technique (SMOTE) and the 90-model SMOTE-variants package in Python to balance the dataset. Due to the multi-class nature of the data and memory constraints, 42 models were successfully used to create a balanced dataset, which was then integrated with a Random Forest algorithm for C-factor classification. The results show a marked improvement in model accuracy across most SMOTE variants, with the Selected Synthetic Minority Over-sampling Technique (Selected_SMOTE) emerging as the best-performing method, achieving an overall accuracy of 0.9524 and a sensitivity of 0.6892. Importantly, the previously observed issue of poor minority class prediction was resolved using the balanced dataset. This study provides a robust solution to the class imbalance issue in C-factor classification, demonstrating the effectiveness of SMOTE variants and the Random Forest algorithm in improving model performance and addressing imbalanced class distributions. The success of Selected_SMOTE underscores the potential of balanced datasets in enhancing machine learning outcomes, particularly in datasets dominated by a majority class. Additionally, by addressing imbalance in LULC classification, this research contributes to Sustainable Development Goal 15, which focuses on the protection, restoration, and sustainable use of terrestrial ecosystems.

Bayesian mesh optimization for graph neural networks to enhance engineering performance prediction

Abstract In engineering design, surrogate models are widely employed to reduce the computational costs of simulations by approximating design variables and geometric parameters from computer-aided design (CAD) models. However, traditional surrogate models often lose critical information when simplified to lower dimensions, and face challenges in handling the complexity of 3D shapes, especially in industrial datasets. To address these limitations, we propose a Bayesian graph neural network (GNN) framework that directly learns geometric features from CAD mesh representations for accurate engineering performance prediction. Our framework leverages Bayesian optimization (BO) to dynamically determine the optimal mesh element size, significantly improving model accuracy while balancing computational efficiency. This approach addresses the challenge of mesh quality by optimizing mesh resolution, ensuring that critical geometric features are preserved in 3D deep-learning-based surrogate models. Unlike existing models that rely on static or predefined mesh resolutions, our method adapts mesh size based on the task, making it highly flexible for a wide range of engineering applications. Experimental results demonstrate that mesh quality directly impacts prediction accuracy, and the proposed BO-EI GNN model outperforms state-of-the-art models such as 3D CNN, SubdivNet, GCN, and GNN in predicting mass, rim stiffness, and disk stiffness. Our method also significantly reduces the computational costs compared to traditional optimization techniques. The proposed framework shows promising potential for application in finite element analysis (FEA) and other mesh-based simulations, enhancing the utility of surrogate models across various engineering fields.

Inverse modeling of electromagnetic emissions using grey wolf optimization

The paper proposes an inverse modeling approach for evaluating radiated electromagnetic compatibility (EMC) in electronic components and circuits. To imitate the electromagnetic behavior of real devices, the suggested methodology generates an equivalent model using elementary dipoles. The Grey Wolf Optimization (GWO) algorithm solves an inverse problem to identify dipole parameters such as locations, intensities, and orientations. This technique improves model accuracy and robustness through a two-phase exploration and exploitation process that allows for effective parameter tuning while reducing simulation time and computational resources. Recent research in near-field measurement techniques illustrates the importance of reliable predictive methodologies in EMC evaluation. This method provides significant information on the impacts of power systems on surrounding electronic circuits, making it a useful tool for EMC testing. It also promotes the development of immunity models for high-frequency applications. Overall, this methodology simplifies the examination procedure, resulting in higher design efficiency for a wide range of electronic systems.

Pseudo-Labeling and Time-Series Data Analysis Model for Device Status Diagnostics in Smart Agriculture

This study proposes an automated data-labeling model that combines a pseudo-labeling algorithm with waveform segmentation based on Long Short-Term Memory (LSTM) to effectively label time-series data in smart agriculture. This model aims to address the inefficiency of manual labeling for large-scale data generated by agricultural systems, enhancing the performance and scalability of predictive models. Our proposed method leverages key features of time-series data to automatically generate labels for new data, thereby improving model accuracy and streamlining data processing. By automating the labeling process, we reduce dependence on manual labeling, which is often labor-intensive and prone to errors in large datasets. This approach enables the efficient preparation of labeled data for applications such as anomaly detection, pattern recognition, and predictive modeling in smart agriculture. Experimental results demonstrate that the automated labeling model achieves 89% accuracy in agricultural environments and reduces data processing time by 30%. Future research will focus on extending the model’s applicability to diverse agricultural settings, enhancing generalization performance, and improving real-time processing capabilities, thereby advancing intelligent and sustainable smart agriculture systems.

Enhancing Lung Cancer Screening with Bidirectional LSTM and GRU Models

Abstract. This study aims to enhance lung cancer patient screening by developing and evaluating bidirectional Long Short-Term Memory (LSTM) and bidirectional Gated Recurrent Unit (GRU) models using the Lung Cancer dataset from Kaggle. The dataset includes 16 features related to patient symptoms and lung cancer status, providing a broad spectrum of symptoms to improve model accuracy. The research advances Artificial Intelligence (AI)-driven healthcare by integrating these sophisticated machine learning techniques into diagnostic processes. The methodology involves four main steps: preprocessing the dataset for model compatibility, defining the model architecture with bidirectional LSTM and GRU layers and evaluating its performance. The results show an overall accuracy of 52.17%, with accuracy, recall, and F1 scores for both cancerous and non-cancerous categories around 50%. Despite the hybrid model's average performance, it establishes a basis for future enhancements. Optimizing model parameters and exploring additional other techniques to improve prediction accuracy and clinical applicability will be done in the future.

Exploring Cutout and Mixup for Robust Human Activity Recognition on Sensor and Skeleton Data

Human Activity Recognition (HAR) is an essential area of research in Artificial Intelligence and Machine Learning, with numerous applications in healthcare, sports science, and smart environments. While several advancements in the field, such as attention-based models and Graph Neural Networks, have made great strides, this work focuses on data augmentation methods that tackle issues like data scarcity and task variability in HAR. In this work, we investigate and expand the use of mixup and cutout data augmentation methods to sensor-based and skeleton-based HAR datasets. These methods were first widely used in Computer Vision and Natural Language Processing. We use both augmentation techniques, customized for time-series and skeletal data, to improve the robustness and performance of HAR models by diversifying the data and overcoming the drawbacks of having limited training data. Specifically, we customize mixup data augmentation for sensor-based datasets and cutout data augmentation for skeleton-based datasets with the goal of improving model accuracy without adding more data. Our results show that using mixup and cutout techniques improves the accuracy and generalization of activity recognition models on both sensor-based and skeleton-based human activity datasets. This work showcases the potential of data augmentation techniques on transformers and Graph Neural Networks by offering a novel method for enhancing time series and skeletal HAR tasks.

The Study on Nondestructive Detection Methods for Internal Quality of Korla Fragrant Pears Based on Near-Infrared Spectroscopy and Machine Learning.

Quality control and grading of Korla fragrant pears significantly impact their commercial value. Rapid and non-destructive detection of soluble solids content (SSC) and firmness is crucial to improving this. This study proposes a method combining near-infrared spectroscopy (NIRS) with machine learning for the rapid, non-destructive detection of SSC and firmness in Korla pears. By analyzing absorbance in the 900-1800 nm range, six preprocessing methods-Savitzky-Golay derivative (SGD), standard normal variate (SNV), multiplicative scatter correction (MSC), Savitzky-Golay smoothing (SGS), vector normalization (VN), and min-max normalization (MMN)-were applied to the raw spectral data. uninformative variable elimination (UVE) and successive projections algorithm (SPA) were then used to extract effective wavelengths. Partial least squares regression (PLSR) models were developed for SSC and firmness based on the extracted data. The results showed that all preprocessing and wavelength-extraction methods improved model accuracy. The optimal SSC prediction model was MSC-SPA-PLSR (R = 0.93, RMSE = 0.195), and the best hardness prediction model was MSC-UVE-PLSR (R = 0.83, RMSE = 0.249). This research aids in establishing a non-destructive testing system, offering producers a rapid and accurate quality assessment tool, and provides the food industry with better production control measures to enhance standardization and market competitiveness of Korla pears.

Modeling heterogeneity of Sudanese hospital stay in neonatal and maternal unit: non-parametric random effect models with Gamma distribution

ObjectiveStudies looking into patient and institutional variables linked to extended hospital stays have arisen as a result of the increased focus on severe maternal morbidity and mortality. Understanding the length of hospitalization of patients after delivery is important to gain insights into when hospitals will reach capacity and to predict corresponding staffing or equipment requirements. In Sudan, the distribution of the length of stay during delivery hospitalizations is heavily skewed, with the average length of stay of 2 to 3 days. This study aimed to investigate the use of non-parametric random effect model with Gamma distributed response for analyzing skewed hospital length of stay data in Sudan in neonatal and maternal unit.MethodsWe applied Gamma regression models with unknown random effects, estimated using the non-parametric maximum likelihood (NPML) technique [5]. The NPML reduces the heterogeneity in the distribution of the response and produce a robust estimation since it does not require any assumptions on the distribution. The same applies to the log–Gamma link that does not require any transformation for the data distribution and it can handle the outliers in the data points. In this study, the models are fitted with and without covariates and compared using AIC and BIC values.ResultsThe findings imply that in the context of health care database investigations, Gamma regression models with non–parametric random effect consistently reduce heterogeneity and improve model accuracy. The generalized linear model with covariates and random effect (k = 4) had the best fit, indicating that Sudanese hospital length of stay data could be classified into four groups with varying average stays influenced by maternal, neonatal, and obstetrics data.ConclusionIdentifying factors contributing to longer stays allows hospitals to implement strategies for improvement. Non-parametric random effect model with Gamma distributed response effectively accounts for unobserved heterogeneity and individual-level variability, leading to more accurate inferences and improved patient care. Including random effects can significantly affect variable significance in statistical models, emphasizing the need to consider unobserved heterogeneity when analyzing data containing potential individual-level variability. The findings emphasise the importance of making robust methodological choices in healthcare research in order to inform accurate policy decisions.

Adaptive robust motion control for hydraulic load sensitive systems considering displacement dynamic compensation

Hydraulic load-sensitive systems (HLSS) are widely used for high power density and energy efficiency. This study introduces an adaptive, energy-efficient HLSS with a valve-controlled variable motor. The system faces challenges from non-linearities, including internal higher-order dynamics due to displacement changes and external unknown disturbances, which hinder precision applications. To address this issue, this study explores HLSS principles to develop an accurate system model. Subsequently, an adaptive robust motion control that considers displacement compensation (DCARC) is proposed using the established model. DCARC can learn unknown parameters online and compensate the model more accurately to improve control accuracy. Experiments show that considering the higher order dynamic effects caused by displacement in the system can improve model accuracy and effectively reduce the burden of parameter adaptation and robust feedback terms. High-precision and energy-efficient HLSS motion is verified and realized in the study. The control accuracy of DCARC is 19.4% higher than that of conventional adaptive robust control (ARC). Under experimental conditions, the proposed system can improve energy efficiency by up to five times compared to valve-controlled fixed displacement motor systems (VFDS).

Deep feature response discriminative calibration

Deep neural networks (DNNs) have numerous applications across various domains. Several optimization techniques, such as ResNet and SENet, have been proposed to improve model accuracy. These techniques improve the model performance by adjusting or calibrating feature responses according to a uniform standard. However, they lack the discriminative calibration for different features, thereby introducing limitations in the model output. Therefore, we propose a method that discriminatively calibrates feature responses. The preliminary experimental results indicate that the neural feature response follows a Gaussian distribution. Consequently, we compute confidence values by employing the Gaussian probability density function, and then integrate these values with the original response values. The objective of this integration is to improve the feature discriminability of the neural feature response. Based on the calibration values, we propose a plugin-based calibration module incorporated into a modified ResNet architecture, termed Response Calibration Networks (ResCNet). Extensive experiments on datasets like CIFAR-10, CIFAR-100, SVHN, and ImageNet demonstrate the effectiveness of the proposed approach.

Data Quality and Data Governance: Investigating the Impact on Data Science Outcomes

Abstract: The purpose of this study is to evaluate the effects that high data quality and effective data governance have on the outcomes of data science, with a particular emphasis on the vital roles that these factors play in improving decision-making within businesses. The rising use of data science projects by enterprises for the purpose of digital transformation presents major hurdles, particularly with regard to the trustworthiness of data and compliance with appropriate legal frameworks. A high data quality improves model accuracy, decision-making efficiency, and error reduction, while good governance boosts compliance with rules, data security, and accessibility, according to the findings of the study, which were derived from an analysis of data collected from professionals working in data science and related sectors. These findings demonstrate the synergy that exists between data quality and governance, underscoring the fact that businesses may obtain more dependable and trustworthy outcomes by prioritizing these components, which in turn supports digital transformation programs that are effective and responsible.

OUCH: Oversampling and Undersampling Cannot Help Improve Accuracy in Our Bayesian Classifiers That Predict Preeclampsia

Unbalanced data can have an impact on the machine learning (ML) algorithms that build predictive models. This manuscript studies the influence of oversampling and undersampling strategies on the learning of the Bayesian classification models that predict the risk of suffering preeclampsia. Given the properties of our dataset, only the oversampling and undersampling methods that operate with numerical and categorical attributes will be taken into consideration. In particular, synthetic minority oversampling techniques for nominal and continuous data (SMOTE-NC), SMOTE—Encoded Nominal and Continuous (SMOTE-ENC), random oversampling examples (ROSE), random undersampling examples (UNDER), and random oversampling techniques (OVER) are considered. According to the results, when balancing the class in the training dataset, the accuracy percentages do not improve. However, in the test dataset, both positive and negative cases of preeclampsia were accurately classified by the models, which were built on a balanced training dataset. In contrast, models built on the imbalanced training dataset were not good at detecting positive cases of preeclampsia. We can conclude that while imbalanced training datasets can be addressed by using oversampling and undersampling techniques before building prediction models, an improvement in model accuracy is not always guaranteed. Despite this, the sensitivity and specificity percentages improve in binary classification problems in most cases, such as the one we are dealing with in this manuscript.

Evaluation of water richness in coal seam roof aquifer based on factor optimization and random forest method.

Water richness evaluation of coal seam roofs is a crucial prerequisite for preventing and controlling water hazards in coal seam roofs. For this purpose, Spearman correlation and GeoDetector were employed for factor optimization to investigate the significance of lithological and structural factors and the impact of interactions among different factors on the water richness of coal seam roofs. Water richness evaluation models of coal seam roofs were separately established via the entropy weight method (EWM), coefficient of variation method (CVM), and random forest method (RFM), and the predictive accuracies of these models were compared. Eleven lithological and structural factors were collected. Through Spearman correlation analysis, 6 factors were identified to have significant influences on water richness. By utilizing the interaction detection of GeoDetector, the effects of interfactor interactions on water richness were explored, and 3 combination factors were selected. AICc was used to determine the model's superiority. The evaluation results of the study area based on factor optimization and three methods were further compared and validated via pumping tests, workface water inflow tests, and three-dimensional high-density electrical method. The results indicated that the RFM exhibited higher prediction accuracy than did the entropy weight and coefficient of variation methods. Additionally, within each method, factor optimization led to improved model accuracy.

Utilizing Artificial Intelligence to Reconcile Indeterminate Lupus Anticoagulant Panel Results

Abstract Lupus anticoagulant (LA) is one laboratory criterion for diagnosing antiphospholipid antibody syndrome (APS), a condition associated with hypercoagulability, pregnancy complications, and death. LA diagnosis requires a complex panel, often performed at a specialized laboratory. Ruling out false positive or transient LAs requires ≥ two positive tests ≥ 12 weeks apart. Most LA tests are negative, with small subsets resulting positive or indeterminate. Indeterminates leave clinicians with limited understanding of appropriate clinical follow-up. Thus, the International Society on Thrombosis and Hemostasis discourages weak resulting nomenclature such as “indeterminate.” Our study used artificial intelligence to predict LA indeterminate results as positive or negative, identifying patients requiring further testing. This IRB approved study reviewed nearly 4000 LA tests performed from 2019-2023. Parameters included CBC, thromboelastography, renal/hepatic function assays, coagulation tests, antiphospholipid antibodies and LA tests. LA test components included screening tests (APTT, Dilute Russel Viper Venom Test (DRVVT) screen), mixing studies (1:1 APTT mix, DRVVT mix), phospholipid-dependent confirmation (DRVVT, hexagonal phase, platelet neutralization), and factor VIII. A positive LA test fulfilled 4 criteria: at least one positive screening test, mixing study and phospholipid-dependent assay and factor VIII inhibitor exclusion. An LA panel meeting 2-3 criteria resulted indeterminate. Studies meeting <2 criteria were negative. This analysis was performed using Machine Intelligence Learning Optimizer (MILO) platform which displays the performance of multiple models. MILO was trained with two classifications: positive and negative cases. Data was split into training (100 training positives; 100 negatives) and testing (primary and secondary validation) sets (100 positives; 1388 negatives). LA panels affected by drug interactions (9.4%) or containing incomplete datasets (21%) were excluded. Sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) determined model efficacy. 2903 negative (84%), 318 positive (9.1%), and 246 indeterminate (7.1%) LA results were included from 3467 patients. Only 84 of the 246 indeterminate patients (34%) were retested. Of these, 28 remained indeterminate, 43 became negative and 13 resulted positive. The best performing MILO models were the K-nearest neighbor (KNN) and logistic regression. The ability of these models to determine which indeterminate cases would become positive or negative was evaluated. KNN displayed the highest specificity (0.88) and PPV (0.57). All models had a high NPV and specificity due to substantive exposure to negative cases for the training and testing. Exclusion of some parameters did not improve the performance of the models. However, even the best model (various KNN models) had low sensitivity and PPV due to a small pool of positive training/testing cases and a much smaller group of indeterminate LA with subsequent positive results. Creating models accurately predicting indeterminate cases that may become positive enables appropriate clinical follow-up. Thus, improving model accuracy will require a larger sample size of positive indeterminates with follow-up.

Uncertainty quantification driven machine learning for improving model accuracy in imbalanced regression tasks

Several factors are known to determine the quality of machine learning models, one of which is the dataset quality. One problem related to the quality of a dataset is the imbalance issue. An imbalanced dataset contains significantly more data points for certain values of the output variable which increases the overfitting risk and negatively affects the prediction accuracy. In this article, we propose using epistemic uncertainty quantification (UQ) of machine learning models to identify rare samples in imbalanced regression problems for balancing the dataset. The developed algorithm, uncertainty quantification-driven imbalanced regression (UQDIR), is guided by UQ to restructure the training set with an adequate weight function using existent samples, eliminating the need for new data collection. After identifying rare samples with UQ, the algorithm selects a sample from the training set, assigns a resampling weight using the new weight function, and finally resamples the selected sample according to its assigned weight. We test UQDIR on several benchmark datasets and different machine learning algorithms, then compare its performance with similar imbalanced regression methods. A metamaterial design problem application is also provided for demonstrating the effectiveness of the algorithm in real-world scenarios. We show that improving the quality of UQ metrics results in improved model accuracy.

Unravelling incipient accidents: a machine learning prediction of incident risks in highway operations

PurposeSafety research has focused on drivers, pedestrians and vehicles, with scarce attention given to highway traffic officers (HTOs). This paper develops a robust prediction model which enables highway safety authorities to predict exclusive incidents occurring on the highway such as incursions and environmental hazards, respond effectively to diverse safety risk incident scenarios and aid in timely safety precautions to minimise HTO incidents.Design/methodology/approachUsing data from a highway incident database, a supervised machine learning method that employs three algorithms [namely Support Vector Machine (SVM), Random Forests (RF) and Naïve Bayes (NB)] was applied, and their performances were comparatively analysed. Three data balancing algorithms were also applied to handle the class imbalance challenge. A five-phase sequential method, which includes (1) data collection, (2) data pre-processing, (3) model selection, (4) data balancing and (5) model evaluation, was implemented.FindingsThe findings indicate that SVM with a polynomial kernel combined with the Synthetic Minority Over-sampling Technique (SMOTE) algorithm is the best model to predict the various incidents, and the Random Under-sampling (RU) algorithm was the most inefficient in improving model accuracy. Weather/visibility, age range and location were the most significant factors in predicting highway incidents.Originality/valueThis is the first study to develop a prediction model for HTOs and utilise an incident database solely dedicated to HTOs to forecast various incident outcomes in highway operations. The prediction model will provide evidence-based information to safety officers to train HTOs on impending risks predicted by the model thereby equipping workers with resilient shocks such as awareness, anticipation and flexibility.

Method for Enhancing AI Accuracy in Pressure Injury Detection Using Real and Synthetic Datasets

Pressure injuries pose significant health risks, especially for the elderly, immobile individuals, and those with sensory impairments. These injuries can rapidly become chronic, making initial diagnosis important. Due to the difficulty of transporting patients from local health facilities to higher-level general hospitals for treatment, it is essential to utilize telemedicine tools, such as chatbots, to ensure rapid initial diagnosis. Recent advances in artificial intelligence have demonstrated potential for medical imaging and disease classification. Ongoing research in the field of dermatological diseases focuses on disease classification. However, the assessment accuracy of artificial intelligence is often limited by unequal class distributions and insufficient dataset quantities. In this study, we aim to enhance the accuracy of artificial intelligence models by generating synthetic datasets. Specifically, we focused on training models for Pressure Injury assessment using both real and synthetic datasets. We used PI data at a domestic medical university. As part of our supplementary research, we established a chatbot system to facilitate the assessment of pressure injuries. Using both constructed and synthetic data, we achieved a top-1 accuracy of 92.03%. The experimental results demonstrate that combining real and synthetic data significantly improves model accuracy. These findings suggest that synthetic datasets can be effectively utilized to address the limitations of small-scale datasets in medical applications. Future research should explore the use of diverse synthetic data generation methods and validate model performance on a variety of datasets to enhance the generalization and robustness of AI models for Pressure Injury assessment.

Performance of the Classical Model in Feature Selection Across Varying Database Sizes of Healthcare Data

Machine learning is increasingly being applied to medical research, particularly in selecting predictive modelling variables. By identifying relevant variables, researchers can improve model accuracy and reliability, leading to better clinical decisions and reduced overfitting. Efficient utilization of resources and the validity of medical research findings depend on selecting the right variables. However, few studies compare the performance of classical and modern methods for selecting characteristics in health datasets, highlighting the need for a critical evaluation to choose the most suitable approach. We analysed the performance of six different variable selection methods, which includes stepwise, forward and backward selection using p-value and AIC, LASSO, and Elastic Net. Health-related surveillance data on behaviors, health status, and medical service usage were used across ten databases, with sizes ranging from 10% to 100%, maintaining consistent outcome proportions. Varying database sizes were utilized to assess their impact on prediction models, as they can significantly influence accuracy, overfitting, generalizability, statistical power, parameter estimation reliability, computational complexity, and variable selection. The stepwise and backward AIC model showed the highest accuracy with an Area under the ROC Curve (AUC) of 0.889. Despite its sparsity, the Lasso and Elastic Net model also performed well. The study also found that binary variables were considered more crucial by the Lasso and Elastic Net model. Importantly, the significance of variables remained consistent across different database sizes. The study shows that no major variations in results between the fitness metric of the model and the number of variables in stepwise and backward p-value models, irrespective of the database's size. LASSO and Elastic Net models surpassed other models throughout various database sizes, and with fewer variables.