Statistics, Optimization & Information Computing

Improved Non-Parametric Double Homogenously Weighted Moving Average Control Chart for Monitoring Changes in Process Location

Olayinka O. Oladipupo — 2025-08-19

Parametric control charts' statistical performance often raises concerns when dealing with processes that lack a predefined probability distribution. In such cases, non-parametric control charts emerge as a viable alternative. Additionally, the adoption of ranked set sampling, with its ability to reduce process parameter variability and enhance control chart performance, proves to be advantageous over traditional simple random sampling techniques. In the field of statistical process control (SPC), control charts are essential tools used to monitor and improve process performance. Among various control charts, the double homogeneously weighted moving average (DHWMA) control chart is recognized for its capability to detect small shifts in process parameters. The study focusses on enhancing the sensitivity and robustness of the NPDHWARSS for detection of shift in process mean and make it more reliable across diverse applications. This study aims to improve the non-parametric double homogeneously weighted moving average control chart, employing the Wilcoxon signed rank test and leveraging the ranked set sampling method, referred to as NPIDHWMA-WSR in this paper. To evaluate the efficiency of the proposed control chart, a comparison was conducted against non-parametric double exponentially weighted moving average using signed rank test (NPRDEWMA-SR) and non-parametric double homogeneously weighted moving average (NPDHWMA) control charts. The comparative analysis highlights the superior performance of the proposed NPIDHWMA-WSR control chart, especially in scenarios involving minimal to moderate changes in process location, as evidenced by metrics such as average run length (ARL), standard deviation run length (SDRL), and median deviation run length (MDRL). Additionally, the study presents a practical application, providing skilled practitioners with tangible evidence of the chart's effectiveness in maintaining both product and process quality. Moreover, the NPIRDHWMA-WSR control chart identified an out-of-control (OOC) condition by the 18th sample, whereas the competing NPDHWMA-RSS control chart didn't signal OOC until the 22nd sample.

The effect of the missing rate and its mechanism on the performance of the imputation methods on different real data sets

Mohammad Mehdi Saber — 2025-09-02

The purpose of this paper is to explore the mechanisms of data missingness and evaluate various imputation techniques used to handle missing data. Missing data is a common issue in data analysis, and its treatment is crucial for accurate modeling and analysis. This paper assesses prevalent imputation methods, including mean imputation, median imputation, K-Nearest Neighbor imputation (KNN), Classification and Regression Trees (CART), and Random Forest (RF). These techniques were chosen for their widespread use and varying levels of complexity and accuracy. Simple methods like mean and median imputation are computationally efficient but may introduce bias, especially when the missingness is not random. In contrast, more advanced methods like KNN, CART, and RF offer better handling of complex missingness patterns by considering relationships among variables. This paper aims to provide guidance for data scientists and analysts in selecting the most appropriate imputation methods based on their data characteristics and analysis objectives. By understanding the strengths and weaknesses of each technique, practitioners can improve the quality and reliability of their analyses.

A Modified Jackknife Liu-Type Estimator for the Gamma Regression Models Data

Ahmed Mutlag Algboory — 2025-07-04

Additional methods were suggested to enhance the biased estimation in the multiple linear regression model. The jackknife-biased estimate approach is essential for addressing high variance and multicollinearity issues. Reduce the effects of multicollinearity with the Liu estimator: This shrinkage method is attractive on several occasions. This document aims to derive a Jackknifed Liu-type Gamma estimator~(JGLTE) and a Modified Jackknifed Liu-type Gamma estimator~(MJGLTE) when multicollinearity exists. Based on Monte Carlo simulations, the proposed estimate outperforms the maximum likelihood estimator (MLE) in terms of mean square error (MSE). Finally, we illustrate the performance of this estimator using real-world data.

A Modified Monsef Distribution Using Quadratic Rank Transformation Map

M.M.E. Abd El-Monsef — 2025-09-04

A new distribution is proposed using the Quadratic Rank Transmutation Map (QRTM), which introduces skew to an initially symmetric base distribution. The Monsef distribution serves as the baseline. Various statistical properties, including moments, the moment-generating function, and the characteristic function, are derived. The parameters are estimated using the maximum likelihood estimation method, and the method’s performance is validated through mean squared errors and average biases. Additionally, two real datasets are used to demonstrate the flexibility of the proposed distribution.

Short Note of Linguistic SuperHypersoft Set

Takaaki Fujita — 2025-09-09

Soft sets provide a mathematical framework for decision-making by associating parameters with subsets of a universal set, effectively managing uncertainty and imprecision [6, 7]. Over time, various extensions of soft sets, including Hypersoft Sets, SuperHypersoft Sets, and Treesoft Sets, have been introduced to address increasingly complex decision-making processes. This paper investigates the Linguistic SuperHyperSoft Set, which is an extension of the Linguistic HyperSoft Set.

Deep Learning for Financial Time Series: Does LSTM Outperform ARIMA and SVR in International Stock Market Predictions?

Abderrahman Yaakoub — 2025-09-09

Time-series analysis and dynamic modeling are crucial in various fields, including business, economics, and finance. This study is based on the prediction of financial time series, which are known for their volatility, nonlinearity, and sensitivity to macroeconomic and psychological factors. This article examines four international stock market indices, such as MASI, S\&P 500, CAC 40, and Nikkei 225, representing Africa, America, Europe, and Asia, respectively, which are challenging to model accurately. This research aims to compare three forecasting models: the classical Autoregressive Integrated Moving Average (ARIMA), the machine learning (ML) model Support Vector Regression (SVR), and the deep learning (DL) model Long-Short-Term Memory (LSTM). The empirical results reveal that LSTM outperforms both SVR and ARIMA in predicting financial time series; SVR outperforms ARIMA in three indices: S\&P 500, CAC 40, and Nikkei 225. In contrast, ARIMA outperforms SVR in the MASI index, proving the effectiveness of this traditional method in specific contexts.

Streamlined Randomized Response Model Designed to Estimate Extremely Confidential Attributes

Ahmad M. Aboalkhair — 2025-07-14

When addressing highly sensitive topics, respondents may provide incomplete or untruthful disclosures, compromising data accuracy. To mitigate this issue, this study introduces an innovative and efficient randomized response framework designed to enhance the estimation of highly sensitive attributes. The proposed model refines Aboalkhair’s (2025) framework, which has been established as an effective alternative to Warner’s and Mangat’s models. This study evaluates the conditions under which the new model achieves greater efficiency than existing approaches. Through theoretical analysis and numerical simulations—accounting for partial truthful reporting—the results demonstrate the model’s superior efficiency. Additionally, the paper quantifies the privacy protection level afforded by the new approach.

A Three-Parameter Alpha Power Erlang Distribution for Modeling Lifetime Data

Maryam Aliabadi — 2025-08-19

The alpha power family of distributions introduces a new type of distribution by adding an extra parameter to its baseline distribution, making it highly suitable for lifetime data analysis. In this paper, we propose the three-parameter alpha power Erlang (APEr) distribution, where the baseline Erlang distribution is composed of the sum of multiple exponential distributions and is a special case of the gamma distribution. After defining the probability density and cumulative distribution functions of the APEr distribution, we present its theoretical properties, including quantiles, moments, and maximum likelihood estimation (MLE) of its parameters.
Furthermore, using a simulation study, we verify the consistency of the parameter estimators and demonstrate the improvement in inference quality with increasing sample size. Finally, we compare the performance of the proposed distribution in fitting on two real-world datasets against the alpha power exponential, exponential, and Erlang distributions, demonstrating its superiority in these applications.

A Generalized Approach to Time-Fuzzy Soft Expert Sets for Decision-Making

Naser Odat — 2025-08-19

As an extension of the fuzzy soft set, Ayman A. Hazaymeh presented the idea of the time-fuzzy soft set in his doctoral thesis in 2013. By introducing the Generalized Time-Fuzzy Soft Expert Set (GT-FSES) as an additional extension of the fuzzy soft set, we expand on this concept in this paper. We examine the most important functions of this new architecture, such as intersection, complement, union, and the logical operations "AND" and "OR." Additionally, we show how GT-FSES may be used practically to solve decision-making issues, providing a fresh method for managing complexity and ambiguity in decision-making processes.

Sharia Stock Return Volatility Model through Bayesian MSGARCH with Asymmetric Effects and Data Structure Changes

Afridho Afnanda — 2025-08-29

The Indonesian Sharia stock market has experienced significant growth over the past year, accompanied by an increase in market capitalization. However, high volatility remains a critical challenge for investors when deciding to invest in Sharia stocks. Modeling Sharia stock return volatility is essential to help investors minimize investment risks. This study aims to identify the most effective model for measuring volatility in the Sharia stock market by comparing classical models such as Generalized Autoregressive Conditional Heteroscedasticity (GARCH), asymmetric models including Exponential GARCH (EGARCH), Threshold GARCH (TGARCH), and Asymmetric Power GARCH (APGARCH), along with the Markov Switching GARCH (MSGARCH) and a Bayesian MSGARCH model that incorporates structural regime changes. The results indicate that the Bayesian MSGARCH model outperforms the other models in capturing the volatility of the Jakarta Islamic Index Sharia stock returns, achieving the lowest prediction error and improved accuracy in parameter estimation. Moreover, the study reveals that investment activities influence the volatility structure during periods of market appreciation and depreciation, with identifiable durations, thereby providing valuable insights for the formulation of effective investment strategies.

Optimal Gene Selection and Machine Learning Framework for Alzheimer’s Disease Prediction Using Transcriptomic Data

Omar Khaled Mohamed El-Shahat — 2025-09-02

Accurate Alzheimer’s Disease (AD) prediction using gene expression data is significantly challenged by ultra-high
dimensionality (38,319 genes) and class imbalance (697 AD vs. 460 controls). To overcome these barriers, we developed
an end-to-end machine learning framework integrating advanced feature engineering with optimized classification. Our
study leveraged 1,157 post-mortem dorsolateral prefrontal cortex samples from multi-cohort repositories, selected for
their established relevance to Alzheimer’s disease (AD) pathology, and employed adaptive linear interpolation with
biological replicates to impute sparse missing values (<5 % per gene) while minimizing noise. We rigorously evaluated
four feature selection approaches: ANOVA F-value filtering (emphasizing inter-group expression differences), Mutual
Information scoring (detecting non-linear gene-AD relationships), L1-SVM regularization (simultaneous sparse selection
and classification), and Correlation-based elimination (reducing feature redundancy). Through exhaustive hyperparameter
tuning (120 configurations), L1-SVM proved optimal, identifying 2,890 biologically coherent genes (including known
Alzheimer’s disease markers APOE, BIN1, and CLU) with 92.5% dimensionality reduction and greater than 99% signal
retention. Eight classifiers were benchmarked on this refined gene set. A support vector machine (SVM) with radial basis
function kernels achieved peak performance: 94.37% accuracy, 96.32% precision, 94.24% recall, and a 95.27% F1-score.
Crucially, the model demonstrated clinical robustness with only 8 false negatives and 5 false positives—exceeding existing
transcriptomic models by ≥ 7% specificity. Validation (1,000 iterations) confirmed stability (F1-score SD: ±0.38%). This
framework enables cost-effective AD screening (reducing genomic testing burden by 92.5%) and provides mechanistic
insights through its interpretable gene panel, advancing precision neurology.

Enhancing Diabetes Disease Prediction and Privacy Preservation via Federated Learning and PSO-WCO Optimization

Alaa A. Almelibari — 2025-08-23

Diabetes mellitus is a leading non-communicable disease, affecting over 537 million individuals globally. Its progression, often influenced by obesity and genetic factors, poses significant health risks, including cardiovascular, renal, and neurological complications. Early detection is essential to minimize these risks. This study addresses class imbalance using Synthetic Minority Over-sampling Technique (SMOTE) and evaluates various classifiers, with AdaBoost achieving the best performance (94.02\% accuracy, 93.32\% F1 score, and 0.95 AUC). To further enhance prediction while preserving data privacy, a novel Federated Learning with Particle Swarm Optimization (FLPSO) model is introduced. In centralized learning, AdaBoost combined with PSO-WCO (Particle Swarm Optimization -Weighted Conglomeration Optimization) attained 96.40\% accuracy, while FLPSO in a federated setup achieved 98.30\%, surpassing existing methods. The proposed model effectively balances prediction accuracy, data privacy, and communication efficiency, highlighting its potential in secure and reliable diabetes prediction and its applicability to related health risk assessments.

Wavelet Daubechies Enhanced Average Chart Incorporating Classical Shewhart and Bayesian Techniques

Hutheyfa Hazem Taha — 2025-09-02

This article aims to improve tools in monitoring processes of production by presenting four new control charts based on the wavelet analysis with the Daubechies wavelet. The proposed charts consist of the classical average chart with approximate coefficients, the Bayesian average chart with approximate coefficients, the classical average chart with detailed coefficients and the Bayesian average chart with detailed coefficients. These charts were used on actual data of body temperatures of newborns in Valia Hospital, Erbil, Kurdistan, Iraq. The proposed charts resist noise because low-pass and high-pass filtering is performed in the wavelet transformation to separate smooth trends from noise. The new charts were evaluated against classical Shewhart average and Bayesian average charts using simulations under control and various mean shift situations. Average Run Length and Control Limit Width, as performance measures, were obtained as the new charts show a better performance than traditional average charts for the case of small to medium size shifts in temperature. This improves the ability to supervise the production process, for example, in medicine by tracking newborns’ temperatures at hospitals.

SBPar scanning: Toward a complete optimal skeleton scan strategy for Additive Manufacturing

Nadir RIHANI — 2025-06-23

In a previous work (Prog Addit Manuf 6:93–118, 2021), a novel Additive Manufacturing scan strategy was designed; the Skeleton-Based Perpendicular (SBP) scanning should show minimal trajectory series compared to classical exiting hatching patterns used in the literature. In contrast, this pattern should lead to mechanical anisotropy due to the one-way oriented printing if it is applied in all part’s layers; a complementary scan strategy must be designed to balance the SBP orientations. This important constraint led the author of this paper to develop the “Skeleton-Based Parallel” (SBPar) strategy as a SBP’s complementary scan for avoiding such issues. Subsequently, the present work details the design of the SBPar pattern and the corresponding scan length; analytical formulations are drawn-up for a simple rectangle as a proof of the concept. Therefore, the superposition of SBP and SBPar constitutes the total skeletal scanning (SB). Results emphasized two conflictual interests: apart from stripe scan, the proposed SBPar scan exhibits a maximized trajectory compared to the other scan strategies; thus, it seems lastly compromising the minimization objective targeted by SBP scan. On the other hand, according to this maximization aspect, the second interest is regarded in terms of surface control which requires maximizing matter spreading and thereafter offering higher densification to the processed surfaces. Furthermore, The SB and the classical scan strategies showed degrees of length-similarities according to decision variables adopted herein. Further works will be dedicated to the implementation of the Skeletal-Based trajectory within real 3D-parts and then to the associated mechanical characterization.

Robust Numerical Approach to CRR Model under Self-financing Assumption

John Abonongo — 2025-09-06

The accurate pricing of options is crucial for minimizing financial risks and making informed investment decisions in dynamic markets. Traditional models like the Black-Scholes often fail to account for the early exercise feature of American options and the self-financing replicating portfolio concept, leading to less realistic pricing. This study address these gaps by employing various metaheuristic algorithms, including Particle Swarm Optimization, Differential Evolution, Grey Wolf Optimization, and Simulated Annealing Algorithm, to estimate the parameters for a modified Cox-Ross-Rubinstein model. We derive a Brownian motion model incorporating upward and downward factors and use the Euler-Maruyama method to simulate stock price paths. By comparing these simulated paths with real stock data, we evaluate the effectiveness of the estimated parameters. Additionally, we improve the numerical method for estimating American option prices via the CRR model by integrating the self-financing replicating portfolio concept. The results demonstrate that Particle Swarm and Grey Wolf optimization algorithms provide parameter estimates that yield simulated paths closely matching the real stock data, thereby offering computationally realistic prices for American options. This study highlights the potential of integrating metaheuristic algorithms with traditional models to enhance the accuracy and reliability of option pricing.

Optimal Placement and Sizing of PV Systems and D-STATCOMs in Medium-Voltage Distribution Networks Using the Atan-Sinc Optimization Algorithm

Oscar Danilo Montoya Giraldo — 2025-07-28

The integration of photovoltaic (PV) systems and D-STATCOMs in medium-voltage distribution networks offers significant potential for enhancing voltage profiles, minimizing power losses, and improving overall system efficiency. This paper introduces the \textit{Atan-Sinc Optimization Algorithm} (ASOA), a novel metaheuristic technique specifically designed for addressing the optimal placement and sizing of PV systems and D-STATCOMs. The ASOA, proposed for the first time in this study, leverages the unique mathematical properties of the $\atan\left(x\right)$ and $\sinc\left(x\right)$ functions to explore and exploit the solution space efficiently. To validate the effectiveness of the ASOA, a comprehensive comparative analysis was conducted against two state-of-the-art methods: the vortex search algorithm (VSA) and the sine cosine algorithm (SCA). Results demonstrate that the ASOA outperforms these benchmark methods in terms of solution quality, convergence rate, and robustness. An economic metric confirms the superior capability of the ASOA in solving this complex optimization problem (reductions of about $35.5429\%$ and $35.6707\%$ for the 33- and 69-bus grids, respectively). This research highlights the ASOA as a promising tool for enhancing the planning and operation of distribution networks, setting a strong foundation for future applications in power system optimization.

An Energy-Efficient Pathfinding Model for Wireless Sensor Networks in IoT Using Whale Optimization Algorithm

Soukaina BOUAROUROU — 2025-07-23

The Internet of Things (IoT) offers the ability of device-to-device seamless connectivity, which enables real-time data collection and collaboration. Wireless Sensor Networks (WSNs), which are collections of geographically dispersed sensor nodes, are integral to IoT systems but suffer from low energy, storage, and wasteful data transmission, causing network instability, latency, and high energy consumption. To address these issues, the current research proposes a novel Pathfinding algorithm based on the Improved Whale Optimization Algorithm (IWOA) for WSNs. The aim of the current research is to enhance the network's performance by optimizing energy consumption, hop count, and data transmission efficiency. The proposed method utilizes intermediate sensors and optimizes the transmission paths step by step with the assistance of IWOA, thus performing efficient energy-saving data routing. The simulation outcomes indicate that the Whale Optimization Algorithm outperforms the Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) approaches with 30% improvement in network lifetime, 10% higher number of active nodes, 15% higher successful packet deliveries, and 17% lower data transmission delay. These results illustrate the effectiveness of the introduced algorithm in maximizing WSN performance and hence are an important contribution to decentralized peer-to-peer and distributed systems.

Energy-Optimized Intelligent Distributed Energy Resources in a Microgrid

Anish Vora — 2025-08-28

The modern smart grid replaces old power networks with networked microgrids with a high penetration rate of energy-storing technology and renewable energy sources. The control strategy is one of the most crucial elements in operating a microgrid power system. Although different control methods have been examined to control hybrid microgrids with interlinking converters, further research is required. A distributed energy system is built on integrating battery energy storage systems (BESS) and renewable energy sources like wind, solar and small hydro systems. The charging facilities for electric cars are also included in this scheme. This work proposed a novel Zebra-based Deep Belief Neural Mechanism (ZbDBNM) with a robust control mechanism. Using a Zebra-based fitness function, this novel approach predicts and optimizes energy cost, Total Harmonic distortion (THD) and power loss to match established norms. An evaluation of the proposed control approach's effectiveness and efficiency against established techniques is provided through comparison.

Comparative Control of PWM-CSCs in Single-Phase Microgrids with Sinusoidal Injection

Angélica Mercedes Nivia-Vargas — 2025-09-10

This paper investigates control strategies for pulse-width modulation current source converters, which play a crucial role in integrating renewable energy sources and managing variable loads in modern power systems. A comparative study is presented between a nonlinear proportional-integral controller and two passivity-based approaches: Interconnection and Damping Assignment Passivity-Based Control and PI-PBC. The analysis aims to evaluate the dynamic response, robustness, and stability of each method under realistic operating conditions. Simulation results reveal that while PI-based controllers perform satisfactorily under steady-state or slowly varying conditions, their performance deteriorates significantly during abrupt load changes or transient disturbances. In contrast, passivity-based strategies demonstrate superior robustness, enhanced disturbance rejection, and improved system stability across a wide range of scenarios. Comprehensive simulations were conducted under variable load profiles, including step and ramp disturbances, to assess control performance using key metrics such as current tracking accuracy, total harmonic distortion, and settling time. The results provide valuable insights for designing reliable and adaptive control schemes in applications such as microgrids, electric drives, and energy conversion systems.

A hybrid Machine Learning approach for air quality prediction in Morocco: combining CatBoost with metaheuristic optimization algorithms

Rachid ED-DAOUDI — 2025-08-20

Air pollution poses serious risks to public health and environmental sustainability, particularly in rapidly
urbanizing areas of developing countries. This study investigates whether combining machine learning algorithms with
metaheuristic optimization techniques can improve the accuracy and efficiency of air quality prediction in Morocco. The
main objective is to compare direct classification of Air Quality Index (AQI) categories with a regression-based approach,
and to evaluate the effectiveness of two optimization strategies—Arithmetic Optimization Algorithm (AOA) and Hunger
Games Search (HGS)—in tuning the CatBoost model’s hyperparameters. Using five months of air quality data from two
monitoring stations in Ait Melloul, we modeled concentrations of PM2.5, PM10, CO, and derived corresponding AQI
classifications. The hybrid approach demonstrated that regression-based classification improved accuracy by nearly 30
percentage points over direct classification. Moreover, HGS achieved similar predictive performance to AOA but was over
twice as computationally efficient. CO concentration predictions in residential areas achieved high accuracy (R2 > 0.95),
while particulate matter predictions revealed limitations in capturing extreme pollution events. These findings suggest that
combining gradient boosting with metaheuristic optimization is a promising strategy for developing scalable and accurate air
quality forecasting systems in North African urban environments, with important implications for public health protection
and environmental policy implementation

Implementation of DRBEM with predictor-corrector for infiltration in four layered heterogeneous soil under impermeable and non-impermeable conditions

Millatuz Zahroh — 2025-09-03

This research focuses on the steady infiltration problem in a furrow irrigation channel consisting of four heterogeneous soil layers (Horizon O, A, B, and C), incorporating the root water uptake process. The rectangular irrigation channel is analyzed under both impermeable and permeable conditions. The problem is formulated as a mathematical model using a nonlinear partial differential equation (PDE). After being transformed into a linear PDE, the equation is determined using the Dual Reciprocity Boundary Element Method (DRBEM) combined with a predictor-corrector scheme. The implementation of DRBEM to solve the model is carried out in two stages. In the first stage, the model is solved without considering root water uptake, and the resulting solution is used to estimate the water pressure response function using the predictor-corrector scheme. In the second stage, DRBEM is applied again with the root water uptake process. The numerical solution obtained represents the suction potential value. The study examines water infiltration characteristics across the four heterogeneous soil layers influenced by root water uptake. The results reveal distinct water movement patterns, with the uppermost layer showing divergent behavior and the lowest layer showing convergent behavior. Additionally, the infiltration in impermeable and permeable channels shows that the permeable condition results in greater water infiltration. These suggest higher water absorption, leading to increased soil water content.

Automated Initialization Method in Convolutional Neural Networks using BO Applied to Plant Disease Classification

Saloua Lagnaoui — 2025-09-04

The Convolutional Neural Network (CNN) stands out as the most effective deep learning model for image classification due to its utilization of convolutional kernels for feature extraction. The initialization methods of these kernels significantly impact a CNN performance, with proper initialization leading to faster convergence and improved overall performance, while poor initialization can hinder the learning process.
Our paper introduces a novel algorithm called BO-IKM, which leverages Bayesian Optimization to determine the optimal kernel initialization methods for each convolutional layer in a CNN. This systematic approach enhances the model's accuracy and precision by identifying the best initializers. We validated the effectiveness of BO-IKM using the "Plant Pathology 2020" dataset, a challenging image classification problem. The results were compelling, showing significant improvements in accuracy and precision for CNN models optimized with BO-IKM compared to those using standard initialization methods. These findings underscore the potential of BO-IKM to enhance CNN performance across various image classification tasks.

A Novel Approach to Solving Singularly Perturbed differential algebraic equations: Regularized Laplace Homotopy Perturbation Method

Khalid Farhan — 2025-07-28

In this paper, we used the Laplace Homotopy Perturbation Method (LHPM) to solve the system of Singularly Perturbed differential algebraic equations (DAEs) with an initial condition. We have added an optimization parameter to LHPM to obtain more accurate solutions . Examples are solved using the method presented in this paper, and the calculated results were compared with the Rung-Kutta and Euler methods to observe the accuracy and efficiency of the proposed method.

Hybrid Deep Learning Technique for Cybersecurity Detection and Classification

Akhila Reddy Yadulla — 2025-08-28

Nowadays, cyber threats (CT) evolve rapidly, and this necessitates developing strong and intelligent prediction models that are effective for the detection and classification of cyber security (CS). Hence, a new Elman Crayfish network (ECFN) is proposed to predict and classify CT. In this study, a Kaggle CS threat dataset is trained with Python to develop a more effective classification model. The dataset undergoes a data refinement stage, where noisy data is preprocessed to improve precision. In order to effectively choose the features, a Crayfish Optimization Algorithm is applied in a spatiotemporal feature analysis to select the relevant attributes that contribute to classification. The ECFN utilizes these chosen features to predict CT more effectively. Finally, the detected attacks are classified, and the performance is measured to obtain high accuracy and reliability in detecting CT. The developed method improves CS protection by optimizing the selection process and improving the accuracy of classification. The model's performance is evaluated with metrics like F score, accuracy, recall, precision, and error rate, and the comparison of the results with existing approaches proves its efficiency.

CAT-VAE: A Cross-Attention Transformer-Enhanced Variational Autoencoder for Improved Image Synthesis

Khadija Rais — 2025-07-13

Deep generative models are increasingly useful in medical image analysis to solve various issues, including class imbalance in classification tasks, motivating the development of multiple methods, where the Variational Autoencoder (VAE) is recognized as one of the most popular image generators. However, the utilization of convolutional layers in VAEs weakens their ability to model global context and long-range dependencies. This paper presents CAT-VAE, a hybrid approach based on VAE and Cross-Attention Transformers (CAT), in which a cross-attention mechanism is employed to promote long-range dependencies and improve the quality of the generated images. On the Ultrasound breast cancer dataset, the CAT-VAE achieved better image quality (FID 8.7659 for Malignant and 7.8761 for Normal) compared to the standard VAE. An experiment was conducted where a CNN classifier model was trained without data augmentation, with augmentation based on VAE, and using synthetic data generated by CAT-VAE. The CNN achieved the highest accuracy (97.00%) when trained with CAT-VAE synthetic images. A classification accuracy of 86.67% was achieved with mixed datasets of real and synthetic images, demonstrating that CAT-VAE improves generalization and resilience. These results highlight CAT-VAE's ability to produce diverse and realistic synthetic datasets.

Galerkin Method for the Solvability of a Micropolar Fluid Flow Model with Novel Frictional Boundary Conditions

El-Hassan Benkhira — 2025-09-02

We investigate a mathematical model describing the flow of an incompressible micropolar fluid within a bounded domain of $\mathbb{R}^3$. The fluid's behavior is governed by a non-symmetric constitutive law, coupled with a couple stress tensor. Frictional boundary conditions are imposed through homogeneous Neumann conditions for the angular velocity field, along with a friction coefficient $h \in L^\infty(\partial\mathcal{O})$, which depends on the tangential component of the velocity field. To address the problem, we derive a variational formulation leading to a coupled system consisting of a variational equation with nonlinear terms governing the velocity field and a linear one describing the microrotational velocity. By applying the Galerkin method, the Cauchy-Lipschitz theorem, and compactness results, we obtain an approximate weak solution to this system.

The Effect of Applying Transfer Learning Approach on Medical and Non-Medical Imaging: Skin Cancer and Flower Types

Farah Alkhalid — 2025-08-21

Transfer Learning is an important technique used
to transfer knowledge from one pre-trained model (called source
domain) to another (called destination domain), this technique
improves good evaluation especially when the dataset of
destination is small, transfer learning may achieve a good
valuation and may not, this is depending on the common
features between source and destination domain. In this work,
an investigation is proposed to show the effects of using Transfer
learning on medical and non-medical images. In this work, three
datasets are used (International Skin Imaging Collaboration
(ISIC), Human Against Machine with 10000 training images
(HAM10000), and Flowers), two for skin cancer lesions as
medical images and the third is flowers types, In addition, four
pre-trained models are used (InceptionVersion3, Residual
neural Network with 50 layers (ResNet50), Mobile network
(MobileNetV2) and Extreme version of Inception (Xception).
The results show that transfer learning does better using
nonmedical images than medical images, and the best pre-model
metrics are got from Xception model, with an accuracy of
approximately 89% in non-medical images and 68% in medical
images, this is because the pre-trained model is fruitful when the
features are common between the source and destination
domain, these common features are more available in
nonmedical than medical (especially in skin lesions).

Pain Intensity Recognition from Facial Expression Using Deep Learning

Nusrat Tania — 2025-08-31

Pain weaves its way into daily life, turning ordinary tasks into challenges and testing our strength in unexpected ways. Pain can be observed in a human’s face and can be understood as pain intensity from patients’ verbal acquisition. However, for non-verbal patients—such as those in the ICU, individuals with mental challenges, or even AI—detecting and interpreting pain remains a complex challenge. To extend this, many researchers have done remarkable research and are still trying to find a solution with acceptable accuracy. This research paper presents a hybrid parallel model to detect the intensity of pain from facial expressions. Following the Prkachin and Solomon Pain Intensity (PSPI) metric, we considered 16 pain levels, which were divided into four subranges. We call these sub ranges as ” No Pain”, ”Mild Pain”, ”Moderate Pain”, and ”Severe Pain”. Our parallel feature fusion model consists of a fully connected network with inputs from two deep CNN models, one being VGG19, and the other model can be ResNet50, DenseNet121, or InceptionV3. Thus, we have 3 parallel feature fusion models (PFFM), respectively, PFFM-1, PFFM-2 and PFFM-3. Besides, we trained and evaluated our models using the McMaster Shoulder Pain dataset, where PFFM-2 emerged as the top performer, achieving an 82.54\% accuracy in assessing pain intensity from facial expressions. By outperforming existing pain detection systems, this breakthrough bridges the gap between human perception and AI, enabling more precise and reliable pain interpretation.

Attention-Augmented EfficientNetV2B0 for Multi-Class Cardiac Disease Classification from Cine MRI

Shivan Hussein Hassan — 2025-09-07

This study presents a pipeline of deep learning for multiclass diagnosis of cardiac disease from cine MRI that combines three significant innovations: an attention-augmented EfficientNetV2B0 backbone for enhanced spatial discrimination of cardiac anatomy, domain-specific preprocessing using CLAHE and best slice selection to enhance prominent myocardial features, and a patient-level ensemble strategy that aggregates slice-wise predictions into robust diagnostic outputs. The model accounts for the volumetric and heterogeneous nature of cardiac MRIs, unlike traditional perslice approaches. We evaluated our system on the MICCAI 2017 ACDC dataset for five classes dilated cardiomyopathy (DCM), hypertrophic cardiomyopathy (HCM), myocardial infarction (MINF), abnormal right ventricle (ARV), and normal (NOR) with a )90.0 %( overall test accuracy and macro F1-score of 0.9013. Performance per class was extremely high in ARV and HCM, with precision and recall of over 0.90. Cross-validation also confirmed the stability of the model with a mean accuracy of 69.6 % ± 1.1 and an F1-score of 67.9% ± 2.0. The model exploited transfer learning with partial fine-tuning as well as attention’s saliency maps, achieving both generalizability and interpretability clinically. By smoothing patient level predictions and regulating model attention toward radiological expectations, the system provides a more consistent and trustworthy diagnosis and has the potential to reduce cardiac triage time by as much as 40 %. The essence of the challenge is still the detection of MINF on the basis of faint tissue biomarkers. Overall, our contribution enhances computational cardiology with a realizable, explainable, and highly accurate method for automatic diagnosis.

Ensemble Learning and K-means Models for Lung and Colon Cancer Classification

Abdelwahid Oubaalla — 2025-08-24

According to World Health Organization (WHO) statistics, cancer remains one of the leading causes of death worldwide. The highest number of cancer-related deaths is caused by lung cancer, with approximately 1.8 million fatalities (18.7%), followed by colorectal cancer, responsible for around 900,000 deaths (9.3%). These death rates are increasing in developing countries, where human, material, and technological resources for early detection are sometimes limited. In Morocco, for example, according to WHO statistics, the mortality rate for lung cancer stands at 21.6%. Diagnosis of histopathological images is one of the most effective ways of confirming or denying the existence of this type of cancer. Traditionally, this analysis is done manually by pathologists, which makes this process time-consuming and the outcome largely depends on the expertise of the pathologist. Automating this process can enable early detection and considerably increase the chances of cure. Thanks to the remarkable results achieved using machine learning techniques, many research projects have attempted to capitalize on these advances and apply them to automate and improve cancer detection accuracy. Despite advancements in deep learning-based classification, achieving consistently high accuracy remains a challenge. In this paper, we propose a new approach that uses the K-means model and gamma correction function to preprocess histopathological images from the LC25000 dataset, and transfer learning and ensemble learning to enhance the classification performance. We have combined two models based on VGG16 and DenseNet pre-trained models. This approach enabled us to achieve an accuracy of 99.96%, which illustrates the importance of combining unsupervised models, transfer learning and ensemble learning to improve the accuracy of histopathological images classification.

Fast and Efficient Feature Selection in AI Application Based on Enhanced Binary Secretary Bird Optimization Algorithm

Amr H. Abdelhaliem — 2025-10-08

Metaheuristic algorithms, which draw inspiration from natural phenomena, have emerged as robust tools within computational intelligence and are widely applied across various fields. The effective use of artificial intelligence requires extracting pertinent information from extensive datasets. Working with big data presents several obstacles, including high dimensionality, duplicate data, and extraneous information. Feature selection techniques aim to reduce complexity by identifying and removing unnecessary attributes, which helps optimize computational resources in terms of both processing time and storage requirements. This paper introduces an enhanced binary variant of the Secretary Bird Optimization Algorithm (SBOA) designed to address feature selection challenges. The SBOA is a recent metaheuristic approach that replicates the survival tactics of secretary birds, specifically their hunting and predator avoidance behaviors. As computational methods, metaheuristic algorithms help solve complex optimization tasks. The proposed EB-SBOA incorporates two key improvements to the original SBOA: a refracted opposition-based learning method during initialization to expand population diversity, and a random replacement mechanism to improve convergence precision. The algorithm's effectiveness was tested using 25 benchmark datasets and compared against six contemporary wrapper-based feature selection techniques. Results demonstrate that EB-SBOA achieves superior performance in three key metrics: classification accuracy, average fitness value, and feature reduction. The findings' statistical validity was confirmed through Wilcoxon rank-sum testing.

Characterization of Demand Profiles for Medium-Voltage Substations in Bogotá Using a Python Interface

María Camila Herrera Briñez — 2025-07-01

This document presents a comprehensive methodology for analyzing operational demand forecast data from XM’s website using a Python script executed in Google Colaboratory (Colab). The script is designed to process text files containing forecast data, which are compressed in a zip file and uploaded to the Colab environment for processing. The key functions of the script include decompressing and organizing files, filtering data based on specified keywords, calculating average power values, and normalizing these averages. The script is structured into several sections: first, it describes various libraries and functions used, and then it presents a main script that implements these functions. Finally, the obtained vectors are utilized to plot normalized active and reactive curves, facilitating the visualization of demand patterns over a 24-hour period. This provides insights into the operational dynamics of Colombia's National Interconnected System (SIN). The analysis focuses on forecast data from May 2023 to May 2024, which includes both reactive and active power. The integration of this script into the data processing workflow aims to streamline the analysis of XM’s operational demand forecasts, supporting more informed decision-making for energy management and planning. By transforming raw forecast data into actionable insights, this approach improves understanding of demand patterns, contributing to the effective management of Colombia’s energy resources.

WebGuard: Enhancing Web Security Through an Integrated Developer Platform

Md. Tanvir Rahman Rafi — 2025-09-15

This research presents the development of an integrated developer platform named ‘WebGuard’. The proposedintegrated platform provides solutions for SQL Injection, Cookie and Session Hijacking, Cross-Site Scripting (XSS),Phishing, Distributed Denial-of-Service (DDoS) attacks, and Malware. This study used input validation by generatingautomated regular expressions to detect SQL injection. In addition, stored procedures, parameterized queries, andcryptography are used to detect SQL injection. This platform used secure session ID generation and encrypted userauthentication to prevent cookie and session hijacking. Here, libsodium is utilized to decrypt user authentication. In thisstudy, the cross-site scripting (XSS) mitigation employs input validation, output encoding, and DOMPurify for advancedsanitization. Distributed Denial-of-Service (DDoS) uses a Content Delivery Network (CDN) inWebguard that contains loadbalancing, rate limiting, and a comprehensive incident response plan. Webguard provided malware detection service byusing file type and size validation and heuristic checks. Furthermore, Phishing attacks are also prevented by the proposedplatform. The proposed platform successfully prevented 92.77% of SQL injection attacks out of 828 samples, and it detected6.16% of the provided samples. Webguard successfully prevented 95.12% of cookie and session hijacking attacks out of 41samples. The platform successfully prevented 90.95%, and detected 7.41% of XSS attacks, out of 243 samples. This platformsuccessfully prevented 81.82% of DDoS attacks out of 11 samples. In phishing detection, Webguard successfully detected92.64% out of 231 samples. Finally, this platform successfully detected 87.88% of malware out of 33 samples. Therefore,WebGuard promotes a safer online environment and makes secure development easier for programmers by combining thesefeatures in one location.

Local Density-Aware Oversampling with Noise Resistance for Imbalanced Data Classification

Yuqing Zhang — 2025-10-15

Imbalanced data classification has become a critical challenge in the field of machine learning. Traditional oversampling approaches often suffer from synthetic sample inaccuracy or aggravated class overlap problems. To better address these challenges, this paper proposes an oversampling method that integrates noise detection with adaptive sample generation. Specifically, we first introduce a noise detection strategy based on the average $k$-nearest neighbor distance, which identifies and removes high-interference noisy samples through local density analysis. Next, we design a weight allocation mechanism that jointly evaluates each instance's boundary risk and generation potential, prioritizing the synthesis of higher-weighted samples. Finally, to better preserve the classification boundary, we incorporate a neighbor class-sensitive coefficient into the sample generation process. Extensive experiments on 17 benchmark datasets demonstrate that the proposed method significantly outperforms well-known oversampling-based approaches, achieving superior classification performance.

Vehicle Routing Problem with Synchronization and Scheduling Constraints of support vehicles

Adil Tahir — 2025-10-20

Many transportation planning processes in real-world applications are complex and require strong cooperation
among various vehicles. When using expensive vehicles, their utilization plays a decisive role in an efficient supply chain. In mining production or civil construction processes, such as mining unloading or road building, the machines are typically mobile, and synchronization between different types of vehicles ensures better use of vehicle fleets, reduces traveled distances, non-productive times, and logistics costs. In this paper, we consider two types of vehicles, called primary and support vehicles. Primary vehicles perform operations and are assisted by at least one support vehicle, with support vehicles scheduled according to a First-Come, First-Served (FCFS) policy. We refer to this practical problem as the vehicle routing problem with synchronization and scheduling constraints of support vehicles. To tackle this problem, we introduce three mixed-integer linear programming models. The first approach involves vehicle routing with synchronization only, breaking each task into several subtasks by duplicating nodes in the graph representation, which produces an equivalent network flow problem. The second model addresses subtasks by adding constraints that determine the assignment of each subtask to a specific primary and support vehicles. The third model incorporates an additional FCFS scheduling constraint for support vehicles. Computational results on 100 real-world instances show that the second model reduces the first model’s computational time by 30%. In contrast, the results of the third model indicate that the FCFS constraint for support vehicles has little impact on solution quality and slightly increases computation time, demonstrating the robustness and practical applicability of the scheduling approach.

Modified Bregman extragradient algorithm for equilibrium problems in Banach spaces

Bochra ZEGHAD — 2025-10-25

This paper introduces a modified Bregman extragradient algorithm designed to solve pseudomonotone equilibrium problems in a real reflexive Banach space. The algorithm guarantees weak convergence under mild assumptions and establishes strong convergence under additional conditions. In our proposed algorithm, we utilize two parameters with the Bregman distance and a non-monotonic step size, which is independent of the Bregman Lipschitz constant, to enhance the algorithm's effectiveness. Furthermore, numerical experiments are conducted to validate the performance of the proposed algorithm, demonstrating significant improvements in efficiency compared to traditional algorithms in similar settings.

The Log-Exponentiated Polynomial G Family: Properties, Characterizations and Risk Analysis under Different Estimation Methods

Nazar Ali Ahmed — 2025-09-26

This work presents a new class of probability distributions termed the Log-Exponentiated Polynomial (LEP) G family. We explore its fundamental properties and provide characterizations. The paper focuses on risk analysis using different estimation methods. The LEP G family offers flexibility for modeling various data types. We derive useful expansions for the new family, these expansions facilitate the calculation of moments and other statistical measures. The model’s parameters are estimated using several methods, including Maximum Likelihood Estimation (MLE). We also employ Cramer-von Mises (CVM), Anderson-Darling (ADE), Right Tail Anderson-Darling (RTADE), and Length Bias Extended (LEADE) estimation techniques. A simulation study evaluates the performance of these estimation methods. Bias, Root Mean Square Error (RMSE), and Anderson-Darling distance metrics are assessed. The LEP Weibull model is applied to insurance claims data for risk measurement. Key Risk Indicators (KRIs) like Value-at-Risk (VaR) and Tail Value-at-Risk (TVaR), Tail Variance (TV), Tail Mean Variance (TMV) and Expected Loss (EL) are calculated. We also analyze artificial data to demonstrate the model’s behavior under controlled conditions. The results highlight the impact of different estimation techniques on risk assessment. The LEP G family proves to be a robust and adaptable framework. It provides a valuable tool for statisticians and actuaries in modeling complex datasets. This work contributes to the advancement of distribution theory and its practical applications.

Alpha Power One-Parameter Weibull Distribution: Its Properties, Simulations and Applications to Real-Life Data

Mohammed EL-arbi Khalfallah — 2025-10-02

In this paper, we introduce a new lifetime distribution called alpha power one-parameter Weibull (APOPW) distribution based on the alpha power transformation method has been defined and studied. Various statistical properties of the newly proposed distribution including moments, moment generating function, quantile function, Rényi and Shannon entropy, stress-strength reliability, mean deviations, and extreme order statistics have been obtained.

Several estimation techniques are studied, including maximum likelihood estimation (MLE), Anderson–Darling (AD), least squares estimation (LSE), Cramér–von Mises (CVM), and maximum product of spacings (MPS). The estimators compared their efficiency based on average absolute bias (BIAS), mean squared error (MSE), and mean absolute relative error (MRE), identifying that MLE as the most robust method across various sample sizes increase.

The efficiency and flexibility of the new distribution are illustrated by analysing two real-live data sets, and compare its goodness-of-fit against several existing lifetime distributions.

An application of nonstandard viscosity iterative methods with s-convexity in the generation of fractals for rational maps

Iqbal Ahmad — 2025-10-25

This paper introduces an application of novel fractal patterns, specifically Julia and Mandelbrot sets, generated by a modified class of complex rational maps in which the traditional constant term is replaced with a logarithmic component. By utilizing nonstandard viscosity iterative methods with s-convexity, we derive enhanced escape criteria that refine existing computational algorithms, thereby enabling the precise visualization of intricate fractal structures as Julia and Mandelbrot sets. Our results demonstrate dynamic transformations in the shape and size of these fractals as key input parameters are adjusted. We believe that the insights garnered from this research will inspire and motivate researchers and enthusiasts deeply engaged in the field of fractal geometry.

A Unified Framework for Generalized Contractions via Simulation Functions in $b-$Metric Spaces with Applications to Nonlinear Analysis

Haitham Qawaqneh — 2025-09-16

This paper introduces a novel framework that unifies Istrătescu-type contractions with simulation functions in the context of $b$ metric spaces. We define a new class of mappings, termed Istrătescu type $\Xi$-contractions, which generalize and extend several well-known contraction types from the literature. Our main result establishes the existence and uniqueness of fixed points for such mappings under mild continuity conditions, providing a unified approach to various fixed point theorems. The flexibility of our framework is demonstrated through several corollaries that recover important classical results as special cases. To illustrate the practical utility of our theoretical developments, we apply our main theorems to prove the existence and uniqueness of solutions for nonlinear fractional differential equations and nonlinear Volterra integral equations. The results presented herein not only advance fixed point theory in generalized metric spaces but also offer powerful tools for analyzing nonlinear problems in applied mathematics and related fields.

An introduction to set-valued fractional linear programming based on the null set concept

Palamanga Lompo — 2025-09-17

Set theory is a generalization of interval theory. However, this theory has shortcomings due to the lack of an inverse element for addition. The concept of null sets was therefore introduced to address this issue. Nonetheless, in set-valued optimization, the use of this concept remains largely insufficient. This article, therefore, introduces a linear fractional set-valued optimization problem, the solution to which is based on the concept of null sets. This concept enables a partial order to be established between sets for simple differences and the Hukuhara difference. On this basis, the notions of optimal and H-optimal solutions have been defined. To solve the proposed set-valued linear fractional optimization problem, it is first transformed into a set-valued linear optimization problem. To make this conversion, we have proposed an adapted version of the Charnes and Cooper method applicable to set-valued linear fractional optimization problems. Subsequently, the obtained set-valued linear optimization problem is transformed into a deterministic linear bi-objective optimization problem using the vectorization technique. To apply a classical method for resolution, the bi-objective problem is converted into a single-objective linear optimization problem using the scalarization technique. Finally, an algorithm has been proposed, and two didactic examples have been solved to better illustrate the steps of the proposed procedure.

Database System and Model for Predicting Risk Level of Flood That Damages Rice Farming in Thailand

Wittawin Susutti — 2025-09-11

Rice is always an important economic crop of Thailand as it is not only the staple in every family of the entire country but it also earns the extremely large income among all the Thai crop exports. However, Thai farmers are considered starve and still have to face many difficulties, in particular, flood problem. The flood problem has destroyed rice farming areas over the past decade until now. Risk and severity assessments mainly contribute the government promptly subsidizing the farmers. In any case, the updated and reliable database systems are the main ingredients to develop the model of these assessments. In this paper, we develop a database system from surveying with 5,000 samples over the whole country. All the raw data has been managed to clean and prepare in order to develop model that is used to predict risk level. The model achieves 87.24 percent accuracy with significant level of 0.05. In addition, the model is able to select variables that have a statistically significant effect on the risk level forecast, and these variables could be used to improve the quality and data structure for developing Web Application (WebApp). The WebApp of our research group for individual risk assessment of the rice farmers has been developed by Javascript to the front end while the back end is run by Python. The WebApp was evaluated satisfactions by over 370 farmers from three public hearings. The satisfaction average scores are over 4 to maximum 5 in all categories.

Testing exponentiality based on Progressively Type-I interval censored data

Huda Qubbaj — 2025-08-13

In this paper we propose non-parametric estimates for the information measure entropy when a progressively Type-I interval censored data is available. Different non-parametric approaches are used for deriving the estimates. Entropy-based tests of exponentiality are proposed. The critical values and the power values of the proposed tests are simulated and studied under various alternatives. Real life data sets are presented and analysed.

A Semi-Analytical Approach to Solving the Black-Scholes Equation via Reproducing Kernel Hilbert Spaces (RKHS)

Erisbey Marin — 2025-09-18

This paper presents a semi–analytical method for solving the Black–Scholes equation by embedding its deterministic and stochastic components into a Reproducing Kernel Hilbert Space (RKHS). The deterministic term is approximated via regularized kernel regression, while the stochastic term is modeled using an autoregressive representation in RKHS. The method is validated on both synthetic geometric Brownian motion trajectories and real adjusted closing prices of Apple Inc. (AAPL), comparing the RKHS approach against the Euler–Maruyama scheme. Results show that the proposed method achieves lower RMSE with fewer anchor points, demonstrating its efficiency and robustness for data–driven financial modeling under uncertainty.

A Generalized Mixture of standard Logistic and skew Logistic distributions: Properties and applications

Partha J. Hazarika — 2025-09-14

In this paper, the generalized mixture of standard logistic and skew logistic is introduced as a new class of distribution. Some important mathematical properties of this novel distribution are discussed along with a graphical presentation of the density function. These properties include moment generating function, $m^{th}$ order moment, mean deviation, characteristic function, entropy, among others. Moreover, a location scale type extension of the proposed distribution is considered, and the maximum likelihood estimation method for this model is presented. To examine the performance of the estimated parameters of the proposed distribution, a simulation study is also conducted using the rejection sampling method. Furthermore, an application using two real-life data sets are also illustrated. Finally, the likelihood ratio test is performed to study the discrepancies between proposed model with their counterparts.

Integrating Climate and Environmental Data with Bayesian Models for Malaria Prediction

Makwelantle Asnath Sehlabana — 2025-09-28

Malaria remains a notable public health challenge in endemic regions, with an estimated 263 million cases and 579,000 malaria-related deaths globally in 2023. Climate and environmental factors, such as temperature, rainfall, and the Normalised Difference Vegetation Index (NDVI), play a crucial role in malaria transmission. While statistical models aid in malaria prediction, Bayesian methods remain underutilised despite their ability to incorporate prior knowledge into predictive models. The major contribution of this study is to develop a Bayesian malaria prediction model incorporating climate and environmental data. Both objective and subjective prior distributions are evaluated to determine their effectiveness in improving model performance. The results indicate that a subjective prior outperforms other priors. Additionally, Ehlanzeni (Mpumalanga), Vhembe and Mopani districts (Limpopo) are identified as high-risk malaria regions. The findings suggest that malaria transmission peaks in summer and autumn, particularly in areas where temperatures during the night range from 12°C-20°C, rainfall is moderate (100–200 mm), and NDVI exceeds 0.6. Malaria risk intensifies following months of accumulated rainfall, creating optimal mosquito breeding conditions. These insights may assist malaria control programmes in developing targeted interventions, such as early warning systems and vector management strategies. Future research will explore Bayesian machine learning for malaria prediction.