Accuracy Performance Research Articles

Application of indoor thermal energy cycle and visual communication simulation scene based on Machine vision and optical image enhancement

Traditional thermal analysis methods have limitations in complex environments, and innovative technologies are urgently needed to improve the accuracy of indoor thermal monitoring and optimization. In this paper, machine vision and optical image enhancement technology are combined to simulate the indoor heat energy cycle process to achieve more efficient heat management and visual communication, so as to improve the energy utilization efficiency and residential comfort of buildings. In this paper, a high-resolution camera is used to capture real-time images of indoor environment, image quality is enhanced by image processing algorithm, and heat distribution characteristics are extracted. The machine learning model is used to analyze the dynamic changes of the thermal energy cycle, and the visual model of the indoor thermal energy cycle is constructed by combining the fluid dynamics theory. The experimental results show that the proposed method can effectively identify and display the key areas of indoor heat circulation and their changing trends. The simulation results show that the optical enhanced images can show different temperature regions more clearly, directly reflect the heat flow situation, and have a high agreement with the actual measurement data. Therefore, the indoor thermal energy cycle simulation method based on machine vision and optical image enhancement not only improves the accuracy and real-time performance of thermal energy analysis, but also provides visual support for indoor environmental management.

Wavelet neural network algorithm for hybrid GA in infrared CO2 gas sensor

As the economy develops and the environmental impact of the greenhouse effect becomes more apparent, the need for precise measurement of specific gas concentrations in the air has become increasingly pressing. Nevertheless, as a representative of greenhouse gases, CO2 gas detectors are susceptible to environmental temperature fluctuations, which impairs the accuracy of detection. To address this issue, the research team innovatively combined the genetic algorithm (GA) and the wavelet neural network (WNN) to develop a solution for the temperature compensation problem of the infrared CO2 gas sensor. The non-dominant sorted genetic algorithm II (NSGA-II) was integrated into the GA to achieve a balance between the accuracy, complexity, and temperature performance of the model through multi-objective optimization. The results showed that compared with other existing models, the GA-WNN model proposed in this study can significantly reduce the difference between the detected values and the actual environmental values under various temperature conditions when processing data. Especially at an ambient temperature of 49 °C, for a true CO2 concentration of 2000 ppm, the detection value processed by the GA-WNN algorithm was 2046 ppm, with a relative error of only 2.3 %, far lower than the 9.8 % of Faster RCNN algorithm and 11.5 % of WNN algorithm. The contribution of the research is the proposal of a novel temperature compensation method that significantly enhances the precision of infrared CO2 gas sensors. This is of paramount importance for enhancing the accuracy of gas detection in environmental monitoring and industrial control.

The Effects of Scales on Modeling of Flow Patterns Over Spillway

Climate changes leading to extreme events when the occurrence and intensity of severe phenomena have impeded the development and rehabilitation of hydraulic infrastructure, such as weirs and spillways according to these new conditions. The scale physical model is one of the most important methods for designing these hydraulic structures. The selection of the model scale is necessary for model operation accuracy and design performance because of the effect of this process on design parameters resulting from the model. Scale effects can be seen through the large effect of some forces on the model that are not effective in reality, such as surface tension and viscosity forces, whose effect can be seen on the model with a relatively small scale, but are not effective on the structure at the actual size. The present study is trying to know how the physical model scales affect flow patterns over the spillway of small dam (less than 15m height) and the extent of this effect. Three physical model scales (1/30, 1/75, and 1/100) were used and compared with the actual size of the structure by numerical model (FLOW 3D). The physical and numerical models were tested with discharge values of (250, 350, 500, 750, and 1000) m3/s to evaluate the rating curve, discharge coefficients, pressure distributions, and energy dissipation. The results show the scale has significantly impacted the results, and it is preferable to avoid small scales. The 1/30 scale aligned more closely with the numerical model (the actual dimensions of the spillway) and strikes a compromise between measurement precision and expense. Therefore; the present study recommended using the scale 1/30 physical model or more for small dam spillway design.

Predictive Analysis of Breast Cancer from Full-Field Digital Mammography Images using Residual Network

Breast cancer has been a significant contributor to cancer-related mortality, but advancements in early detection through regular mammography and improvements in treatment modalities have contributed to declining mortality rates in several regions. This study presents a novel approach to cancer diagnosis utilizing Full-Field Digital Mammography images through predictive analysis methods. By using predictive analytic techniques and mammography images, this study offers a novel way to cancer detection. The research involves the application of deep learning techniques to extract valuable insights from cancer images captured by mammography devices. The CBIS-DDSM (Curated Breast Imaging Subset of Digital Database for Screening Mammography) dataset including images from patients with varying types and stages of cancer, is collected and pre-processed to ensure uniformity and quality. Relevant features, including color, texture, and shape characteristics, are extracted, and a rigorous feature selection process is employed to identify discriminative markers. The Residual Network (ResNet) model is selected and trained on the dataset, with a focus on classification accuracy and robust predictive performance. Validation metrics, such as accuracy, IoU (Intersection over Union) score, dice score, and ROC (Receiver Operating Characteristic) curve are employed to evaluate the model’s efficiency. After analysis, the proposed method had the best degree of mass lesion detection accuracy, at 99.24%. This research contributes to the advancement of non-invasive and efficient diagnostic tools, potentially enhancing early detection and intervention in cancer patients. The proposed method not only demonstrates promising results in terms of diagnostic accuracy but also emphasizes interpretability, seamless integration into clinical workflows, and adherence to ethical standards.

The physical information LSTM surrogate model for establishing a digital twin model of reciprocating air compressors

Reciprocating air compressors play a crucial role in industrial production processes. However, due to the complex structure and long operating time of reciprocating air compressors, real-time monitoring to grasp the operating status of reciprocating air compressors has become particularly important. Digital twin is a technology that can reflect the behavior of physical entities in real time, accurately predicting and evaluating the operation status of reciprocating air compressors. However, the establishment of a digital twin model for reciprocating air compressors requires a significant amount of computational resources, which can result in the inability to meet the requirements of real-time performance evaluation. To overcome this limitation, this paper proposes a method for constructing a digital twin model of reciprocating air compressors based on a surrogate model. The surrogate model is constructed based on a long short-term memory neural network with physical information(PILSTM). This model can accurately describe the changes in cylinder pressure by combining physical information. According to the characteristics of cylinder pressure changes, regularization formulas are added to ensure the smoothness of the predicted pressure. The experimental results show that the digital twin model based on the surrogate model has high prediction accuracy and real-time performance. Therefore, this model provides a new method for monitoring the operating status of reciprocating air compressors.

Detection of walnut internal quality via improved YOLOv5 and x‐ray imaging

AbstractThis study proposes a walnut target detection algorithm based on improved YOLOv5 and x‐ray imaging to meet the demand for internal quality detection and removal in the Xinjiang walnut industry. By replacing the C3 module in the backbone layer with the C2f module and the couple‐head in the head layer with the decouple‐head, the algorithm reduces computational complexity, enhances robustness and generalizability, and retains more spatial information, thereby improving the performance of multicategory small target detection. In addition, this paper replaces the original CIOU loss function with the EIOU loss function to improve the convergence speed of the algorithm's accuracy and boundary aspect ratio. Compared with the original model, the improved model, improved YOLOv5, maintains the same average precision for normal walnuts while increasing the average precision for shriveled walnuts and empty‐shell walnuts by 8.2% and 0.4%, respectively. Compared with other mainstream models, such as VGG16, ResNet50, YOLOv5s, YOLOv7, YOLOv8s, YOLOv9s, and YOLOv10s, this model achieves the highest detection accuracy and good detection performance, with a single‐image detection time of 11.9 ms, meeting the requirements for real‐time detection. This work lays a foundation for automatic robot detection of the internal quality and removal of walnuts, showing practical application potential.Practical applicationsXinjiang stands as a prominent producer of walnuts; however, due to the decentralized nature of its cultivation and management processes, a notable prevalence of internal empty shells and shrinkage is observed, which substantially diminishes their commercial value. The integration of YOLOv5 with x‐ray imaging technology promises to enhance the precision of internal quality assessments of walnuts. The improved YOLOv5 model, as delineated in this study, exhibits the highest detection accuracy when benchmarked against a multitude of other models. It achieves a detection latency of merely 11.9 ms per image, thereby satisfying the stringent demands for real‐time detection applications. This model is designed for integration into x‐ray systems with an adjunctive sorting mechanism, which facilitates the inspection and exclusion of defective walnuts on conveyor belts.

Wavelet-based spatiotemporal sparse quaternion dictionary learning for reconstruction of multi-channel vibration data

Accurate reconstruction of multi-source data information plays an essential role in remote intelligent operation, and prognostic and health maintenance (RIO-PHM) of industrial systems. However, traditional signal reconstruction methods such as compressed sensing fail to capture the spatio-temporal characteristics and coupling nature of the multi-source or multi-channel data. To alleviate this bottleneck, in this paper, a new wavelet-based spatiotemporal sparse quaternion dictionary learning (WSTS-QDL) algorithm is formulated for the reconstruction of multi-channel vibration data for the first time. Specifically, the wavelet-based spatiotemporal sparse low-rank matrix (SLRM) algorithm is elaborated and the sparse coefficient matrix of the multi-channel signals can be estimated using the split Bregman iteration (SBI) technique. Then, quaternion-based dictionary learning is introduced for multi-channel data reconstruction according to the updated sparse coefficient matrix during quaternion dictionary learning. Eventually, two equipmental scenarios including run-to-failure data of rolling bearing in bearing test bench and vibration signal of the failured bearing in corn thresher, are utilized for verification. The experimental results demonstrate that the highest recovery accuracy performance with scoring function and the lowest errors are achieved by the proposed method in contrast with the state-of-the-art benchmarks such as orthogonal matching pursuit (OMP) method and spatiotemporal sparse Bayesian learning (SSBL), and the waveform of phase space trajectory and points cloud distribution of the Poincare section in the original signal are similar/same to that obtained by the proposed method.

Gut metagenome-derived image augmentation and deep learning improve prediction accuracy of metabolic disease classification.

In recent years, statistics and machine learning methods have been widely used to analyze the relationship between human gut microbial metagenome and metabolic diseases, which is of great significance for the functional annotation and development of microbial communities. In this study, we proposed a new and scalable framework for image enhancement and deep learning of gut metagenome, which could be used in the classification of human metabolic diseases. Each data sample in three representative human gut metagenome datasets was transformed into image and enhanced, and put into the machine learning models of logistic regression (LR), support vector machine (SVM), Bayesian network (BN) and random forest (RF), and the deep learning models of multilayer perceptron (MLP) and convolutional neural network (CNN). The accuracy performance of the overall evaluation model for disease prediction was verified by accuracy (A), accuracy (P), recall (R), F1 score (F1), area under ROC curve (AUC) and 10 fold cross-validation. The results showed that the overall performance of MLP model was better than that of CNN, LR, SVM, BN, RF and PopPhy-CNN, and the performance of MLP and CNN models was further improved after data enhancement (random rotation and adding salt-and-pepper noise). The accuracy of MLP model in disease prediction was further improved by 4%-11%, F1 by 1%-6% and AUC by 5%-10%. The above results showed that human gut metagenome image enhancement and deep learning could accurately extract microbial characteristics and effectively predict the host disease phenotype. The source code and datasets used in this study can be publicly accessed in https://github.com/HuaXWu/GM_ML_Classification.git.

Parkinson’s Disease Prediction Using Convolutional Neural Networks and Hand-Drawn Image Analysis

Parkinson’s Disease (PD) is a serious neurodegenerative disorder, with over 10 million cases globally in 2020, significantly affecting patients’ quality of life. The progression of the disease has been statistically proven, underscoring the importance of early diagnosis. As many as 80% of those diagnosed with PD begin to experience spinal degeneration, leading to other impairments and disabilities within approximately 10 years. Moreover, up to 35% of patients require assistance to walk or perform daily living activities within 5 years of diagnosis. The proposed study employs a neural convolutional network (CNN) to predict PD using 64x64 pixel hand-drawn images from 244 PD patients and 228 healthy individuals. K-nearest neighbors (KNN)-based feature extraction was applied as a data pre-processing method before feeding the data into the CNN layers. Model training involved tuning hyper parameters and testing several learning rates, ranging from 0.1 to 0.00001. The highest learning rate of 0.001 yielded the best performance, achieving classification accuracies, precision, sensitivity, and F1 score of 97.93%, 92%, 80%, and 86%, respectively, with a 5% increase in performance accuracy. These results demonstrate the model’s effective ability to discriminate between healthy individuals and PD patients based on hand-drawn samples.

Deep learning detects subtle facial expressions in a multilevel society primate.

Facial expressions in nonhuman primates are complex processes involving psychological, emotional, and physiological factors, and may use subtle signals to communicate significant information. However, uncertainty surrounds the functional significance of subtle facial expressions in animals. Using artificial intelligence (AI), this study found that nonhuman primates exhibit subtle facial expressions that are undetectable by human observers. We focused on the golden snub-nosed monkeys (Rhinopithecus roxellana), a primate species with a multilevel society. We collected 3427 front-facing images of monkeys from 275 video clips captured in both wild and laboratory settings. Three deep learning models, EfficientNet, RepMLP, and Tokens-To-Token ViT, were utilized for AI recognition. To compare the accuracy of human performance, two groups were recruited: one with prior animal observation experience and one without any such experience. The results showed human observers to correctly detect facial expressions (32.1% for inexperienced humans and 45.0% for experienced humans on average with a chance level of 33%). In contrast, the AI deep learning models achieved significantly higher accuracy rates. The best-performing model achieved an accuracy of 94.5%. Our results provide evidence that golden snub-nosed monkeys exhibit subtle facial expressions. The results further our understanding of animal facial expressions and also how such modes of communication may contribute to the origin of complex primate social systems.

Development and testing of hybrid (PNM–CFD) mathematical model and numerical algorithm for description of fluid flows in three-dimensional digital core models

Numerical simulation of fluid flow in porous and fractured rocks is an important task for many industrial applications. The most common approaches to constructing such models are direct (CFD or lattice Boltzmann) and porous network (PNM) modeling. Each approach has its own advantages and disadvantages. This paper presents a hybrid mathematical PNM–CFD model for describing fluid flows in three-dimensional digital core models, which has the high speed performance of PNM approaches and high accuracy of CFD models. Numerical technique has been developed for describing fluid flows in three-dimensional digital core models using a hybrid PNM–CFD model. The numerical technique links a one-dimensional pore network solver and a three-dimensional CFD solver into a combined model in original way by constructing a single pressure field for the entire computation domain. To validate the model, several tests were performed, including flow in straight channels and numerical simulation of fluid flow in a microfluidic chip. The test results have shown the adequacy of the hybrid model performance. The hybrid model for determining pressure drop in a branched network with three-dimensional chambers has an error of no more than 5 % when compared to experimental data. Similarly, the error in calculating velocity does not exceed 7 % when compared to the full three-dimensional calculation. The hybrid model has shown an almost twofold increase in calculation speed compared to the full three-dimensional model.

Deep Learning-Based Classification of Powerlifting Movements Using Mediapipe

The analysis of sports movements is of great importance for optimizing sports performance, minimizing injury risks, and ensuring that athletes work with correct techniques. Powerlifting is a power sport consisting of fundamental movements such as bench press, deadlift, and squat. These movements are inherently complex and challenging to execute. Therefore, it is of great importance to perform these movements with the correct technique and safely. The aim of this study was to classify these movements using deep learning methods to ensure that the basic movements in powerlifting sports (bench press, deadlift, and squat) are applied with correct techniques and to minimize the risk of injury. In this study, feature extraction was performed on powerlifting movements using the deep learning-based SqueezeNet model, followed by classification using machine learning methods. The dataset was compiled from 876 images of bench press, deadlift, and squat movements sourced from various online platforms. Additionally, the dataset was expanded through data augmentation techniques, and key points of posture estimation were added to the images using the Mediapipe library. The obtained datasets were classified using Neural Network, Logistic Regression, Support Vector Machine and Random Forest algorithms and model performances were evaluated using various metrics. The findings revealed that the Neural Network model demonstrated superior performance, achieving the highest accuracy (0.989). Additionally, the integration of pose estimation and data augmentation techniques significantly enhanced classification accuracy and overall model performance. The findings of this study show that deep learning methods are powerful tools in sports movement analysis and can make significant contributions to the evaluation of athletes' performance.

Speech Perception and Phonological Representation in 5 to 6 Year-Old Typically Developing Children

Objectives: This study aimed to examine the development of speech perception and phonological representation in typically developing children aged 5-6 years. It also explored the relationship between receptive vocabulary, consonant accuracy, nonword repetition performance, and speech perception and phonological representation. Methods: The study involved 77 typically developing children aged 5-6 years. The speech perception identification tasks included minimal pairs that reflected developmental error patterns. The phonological representation judgment tasks involved manipulating consonants and vowels, as well as the number of distinctive features, to examine performance differences under various speech sound manipulation conditions. Results: Performance on both speech perception and phonological representation tasks significantly increased with age. There were also significant differences in performance based on speech sound manipulation conditions. Specifically, in the speech perception task, performance was significantly higher when items focused on the stopping of fricatives rather than the fronting of velars. In the phonological representation task, performance was significantly better in vowel manipulation compared to consonant manipulation, and with two distinctive feature manipulations compared to one. Speech perception tasks showed a significant correlation only with receptive vocabulary, whereas phonological representation tasks showed significant positive correlations with receptive vocabulary, consonant accuracy, and nonword repetition performance. Conclusion: The study confirms that speech perception and phonological representation are developing between ages 5 and 6 in typically developing children. It also suggests that speech perception and phonological representation develop through a complex interplay with speech production abilities and receptive vocabulary.

Uncovering Tourist Visit Intentions on Social Media through Sentence Transformers

The problem of understanding and predicting tourist behavior in choosing their destinations is a long-standing one. The first step in the process is to understand users’ intention to visit a country, which may later translate into an actual visit. Would-be tourists may express their intention to visit a destination on social media. Being able to predict their intention may be useful for targeted promotion campaigns. In this paper, we propose an algorithm to predict visit (or revisit) intentions based on the texts in posts on social media. The algorithm relies on a neural network sentence-transformer architecture using optimized embedding and a logistic classifier. Employing two real labeled datasets from Twitter (now X) for training, the algorithm achieved 90% accuracy and balanced performances over the two classes (visit intention vs. no-visit intention). The algorithm was capable of predicting intentions to visit with high accuracy, even when fed with very imbalanced datasets, where the posts showing the intention to visit were an extremely small minority.

Information hiding using approximate POSIT representation

AbstractPOSIT is a numerical system consists of sign, regime, exponential, and fraction bits to overcome some limitations of the IEEE‐754 floating point (FP) representation. This work proposes a technique to hide critical information in the least significant bits (LSBs) of the POSIT FP representation by exploiting approximate computing (AC). The proposed technique, called information hiding using POSIT and LSB (IHUPL), explores the opportunities offered by both the POSIT representation and the AC to hide information with a minimum loss in accuracy and other performance metrics. IHUPL offers two options for hiding information: either by embedding one digit of each character into each pixel or by embedding all the digits of the character at the same time into each pixel of the alpha‐red, green, blue, or black/white image. Experimental results are evaluated for benchmark images and showed that IHUPL enhanced the accuracy of embedding data into LSB of POSIT FP by an average of 5%, the image quality improvement rates of IHUPL are 19% and 16% for options 1 and 2, respectively. Besides the encoding method using IHUPL, the paper outlines an extraction decoding technique that saves the original replacement bits in a key‐image to recover the hidden security message.

DCNFYOLO: Dual-Convolution Network and Feature Fusion for High-Precision Smoke Detection

Fast, real-time, and accurate detection of smoke characteristics in the early stage of a fire is crucial for reducing fire losses. Existing smoke detection methods mainly rely on traditional algorithms and smoke sensors, and these approaches have limitations in false detection rates, accuracy, and real-time performance. Therefore, a novel DCNFYOLO network for smoke detection is proposed in this paper. Firstly, Switchable Atrous Convolution (SAConv) is introduced in the YOLOv5 backbone network to enhance the fusion extraction of smoke features by the Convolutional Neural Network (CNN). Secondly, both Distribution Shifts Convolution (DSConv) operator and Efficient Channel Attention (ECA) mechanisms are considered in the neck part to reduce the computational load of the model, and better capture the relationship between channels to improve the detection performance. Finally, to make low-quality examples less harmful to the gradients, the Wise-IoU (WIoU) loss function in the prediction part is used for reducing the competitiveness of high-quality anchor frames during model training, allowing the model to converge more quickly and stably. Experimental results show that the DCNFYOLO network can achieve a remarkable detection accuracy of 96.6%, which has a substantial improvement of 7.7% compared with the original YOLOv5 network performance, thereby validating the effectiveness of the proposed network.

Efficient nonlinear model predictive and active disturbance rejection control for trajectory tracking of unmanned vehicles

This article proposes an efficient trajectory tracking control strategy of unmanned vehicles. The method is based on nonlinear model predictive control (NMPC) and active disturbance reject control (ADRC). The designed control algorithm considers three challenges including nonlinear characteristics, multiple constraints, and external disturbance. First, NMPC method is presented for the nonlinear vehicle model with multiple constraints. To relax inequality constraints and reduce the heavy calculation burden, the penalty term with the variable factor is added to the cost function, an improved continuous/generalized minimum residual method is proposed to solve NMPC online optimization problem. Then, an ADRC scheme is designed to estimate the unknown disturbance via extended state observer, and compensate them by feedback control law in real time, the corresponding parameters are obtained by a novel particle swarm optimization algorithm to further improve the control precision. It ensures the stability to a certain extent. Finally, the results indicate the designed algorithm can raise the calculation efficiency and meet the real-time requirements, and obviously increase the tracking accuracy and robustness performance.

Lightweight monocular depth estimation using a fusion-improved transformer

The existing deep estimation networks often overlook the issue of computational efficiency while pursuing high accuracy. This paper proposes a lightweight self-supervised network that combines convolutional neural networks (CNN) and Transformers as the feature extraction and encoding layers for images, enabling the network to capture both local geometric and global semantic features for depth estimation. First, depth-separable convolution is used to construct a dilated convolution residual module based on a shallow network to improve the shallow CNN feature extraction receptive field. In the transformer, a multidepth separable convolution head transposed attention module is proposed to reduce the computational burden of spatial self-attention. In the feedforward network, a two-step gating mechanism is proposed to improve the nonlinear representation ability of the feedforward network. Finally, the CNN and transformer are integrated to implement a depth estimation network with a local-global context interaction function. Compared with other lightweight models, this model has fewer model parameters and higher estimation accuracy. It also has better generalizability for different outdoor datasets. Additionally, the inference speed can reach 87 FPS, achieving better real-time performance and accounting for both inference speed and estimation accuracy.

Eyes and hand are both reliable at localizing somatosensory targets.

Body representations (BR) for action are critical to perform accurate movements. Yet, behavioral measures suggest that BR are distorted even in healthy people. However, the upper limb has mostly been used as a probe so far, making difficult to decide whether BR are truly distorted or whether this depends on the effector used as a readout. Here, we aimed to assess in healthy humans the accuracy of the eye and hand effectors in localizing somatosensory targets, to determine whether they may probe BR similarly. Twenty-six participants completed two localization tasks in which they had to localize an unseen target (proprioceptive or tactile) with either their eyes or hand. Linear mixed model revealed in both tasks a larger horizontal (but not vertical) localization error for the ocular than for the manual localization performance. However, despite better hand mean accuracy, manual and ocular localization performance positively correlated to each other in both tasks. Moreover, target position also affected localization performance for both eye and hand responses: accuracy was higher for the more flexed position of the elbow in the proprioceptive task and for the thumb than for the index finger in the tactile task, thus confirming previous results of better performance for the thumb. These findings indicate that the hand seems to beat the eyes along the horizontal axis when localizing somatosensory targets, but the localization patterns revealed by the two effectors seemed to be related and characterized by the same target effect, opening the way to assess BR with the eyes when upper limb motor control is disturbed.

Predicting diabetes in adults: identifying important features in unbalanced data over a 5-year cohort study using machine learning algorithm

BackgroundImbalanced datasets pose significant challenges in predictive modeling, leading to biased outcomes and reduced model reliability. This study addresses data imbalance in diabetes prediction using machine learning techniques. Utilizing data from the Fasa Adult Cohort Study (FACS) with a 5-year follow-up of 10,000 participants, we developed predictive models for Type 2 diabetes.MethodsWe employed various data-level and algorithm-level interventions, including SMOTE, ADASYN, SMOTEENN, Random Over Sampling and KMeansSMOTE, paired with Random Forest, Gradient Boosting, Decision Tree and Multi-Layer Perceptron (MLP) classifier. We evaluated model performance using F1 score, AUC, and G-means—metrics chosen to provide a comprehensive assessment of model accuracy, discrimination ability, and overall balance in performance, particularly in the context of imbalanced datasets.Resultsour study uncovered key factors influencing diabetes risk and evaluated the performance of various machine learning models. Feature importance analysis revealed that the most influential predictors of diabetes differ between males and females. For females, the most important factors are triglyceride (TG), basal metabolic rate (BMR), and total cholesterol (CHOL), whereas for males, the key predictors are body Mass Index (BMI), serum glutamate Oxaloacetate Transaminase (SGOT), and Gamma-Glutamyl (GGT). Across the entire dataset, BMI remains the most important variable, followed by SGOT, BMR, and energy intake. These insights suggest that gender-specific risk profiles should be considered in diabetes prevention and management strategies. In terms of model performance, our results show that ADASYN with MLP classifier achieved an F1 score of 82.17 ± 3.38, AUC of 89.61 ± 2.09, and G-means of 89.15 ± 2.31. SMOTE with MLP followed closely with an F1 score of 79.85 ± 3.91, AUC of 89.7 ± 2.54, and G-means of 89.31 ± 2.78. The SMOTEENN with Random Forest combination achieved an F1 score of 78.27 ± 1.54, AUC of 87.18 ± 1.12, and G-means of 86.47 ± 1.28.ConclusionThese combinations effectively address class imbalance, improving the accuracy and reliability of diabetes predictions. The findings highlight the importance of using appropriate data-balancing techniques in medical data analysis.