Company X, a leader in the Life Sciences supply chain, specializes in producing high-quality medical and laboratory equipment. Given the dynamic market influenced by trends, seasonal variations, economic shifts, and global events such as pandemics, traditional forecasting methods often fall short in accuracy. This work seeks to enhance demand forecasting for highly demanded products by implementing advanced time series models and developing a comprehensive planning model that integrates these methods. Utilizing the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework, the work systematically approaches the data-driven challenges of demand and supply management. Each product time series is first statistically classified to determine the presence of seasonality, guiding the selection of appropriate forecasting models. For non-seasonal products, ARIMA, Holt’s Linear, ARIMA-SVR, and Facebook Prophet were evaluated, while seasonal series employed SARIMAX, Holt-Winters, SARIMAX-SVR, and Prophet models. The results indicate that the ARIMA-SVR and SARIMAX-SVR models consistently out-perform others, reducing forecasting errors by approximately 35.3
Modern deep learning optimization faces persistent challenges in computational efficiency and convergence stability due to complex loss landscapes and heterogeneous parameter dynamics across architectures like CNNs, SNNs, and Transformers. While existing adaptive optimizers (e.g., AdamW, Lion) provide partial solutions, they inadequately address inter-parameter dependencies and neural adaptation mechanisms. Inspired by biological principles of synaptic plasticity and cortical organization, we propose PANDA (Parameter Adaptation through Neural Dynamics Alignment), a novel optimization framework that enhances convergence stability and model generalization through three key innovations: 1) Gradient similarity graphs modeling parameter interdependencies; 2) Dynamic importance quantification enabling context-aware parameter updates; and 3) Region-based grouping balancing local-global optimization dynamics. Comprehensive evaluations demonstrate PANDA's superiority over state-of-the-art methods, achieving 73.3
This paper proposes an unpaired shadow removal method based on the CycleGAN framework, in which the generator is structurally enhanced by introducing skip connections. Existing CycleGAN-based methods commonly rely on plain encoder-decoder generators originally tailored for style transfer, which often fail to preserve pixel-level spatial alignment and low-level structural details. This limitation leads to incomplete shadow removal, texture degradation, and inconsistent illumination in the restored images. To overcome these issues, the proposed generator introduces skip connections that inject low-level features directly from the encoder to the decoder. This architectural design helps preserve structural integrity and improve visual coherence across the entire image. By leveraging both global and local information, the network achieves a more accurate reconstruction of shadow-free images. The method is evaluated on the public ISTD dataset. Quantitative results show that it consistently outperforms existing unpaired models including CycG, Mask-CycG, and DC-CycG, measured by root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM). Visual comparisons further confirm fewer artifacts, more complete shadow removal, and more consistent illumination in the restored regions. Feature map analysis further confirms that the skip-connected architecture captures richer and more diverse semantic features, facilitating better decoding. These findings highlight the practical and theoretical value of enhancing CycleGAN-based generator with skip connections, offering an effective solution for high-quality unpaired shadow removal.
This study evaluates the sediment control performance of two sewage pump stations (PS) configurations, PS-1 and PS-2, in Langat Kajang, Malaysia, under varying rainfall conditions. The primary focus is on the removal efficiencies of Total Suspended Solids (TSS) and Volatile Suspended Solids (VSS). Over 100 days (24 July 2023–31 October 2023), operational data were analyzed using Principal Component Analysis (PCA) and Agglomerative Hierarchical Clustering (AHC). PCA results revealed that the first two principal components accounted for 77.83
Accurate river water level prediction is vital for early flood warning and effective water resource management. The traditional methods of river water level prediction based on Autoregressive Integrated Moving Average (ARIMA) and linear regression struggled to capture the nonlinear and dynamic nature of hydrological systems. Therefore, this study evaluates the performance of deep learning models, namely Long Short-Term Memory (LSTM), Basic LSTM, Stacked LSTM, Bidirectional LSTM, and LSTM with Dropout, alongside classical models such as Linear Regression, Moving Average, and ARIMA across five river monitoring stations. Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), and R-squared (R2) metrics on test data. The results consistently showed that the LSTM with Dropout model outperformed all other approaches in terms of predictive accuracy and generalization, achieving the lowest RMSE and MAE values and the highest R2 scores at four out of five stations. The standard LSTM model achieved the best performance at one station but was generally outperformed by its dropout-regularized counterpart. In contrast, classical models exhibited significantly lower performance, particularly in capturing temporal dependencies, with R2 values often below zero. These findings validated the effectiveness and robustness of LSTM-based deep learning, particularly with dropout regularization, in modeling and forecasting river water level dynamics across diverse hydrological conditions. The study supports the integration of advanced machine learning models in river monitoring systems to enhance flood preparedness and mitigation strategies.
Extreme hot temperatures in Malaysia due to climate change has increase the risk of heat-related health issues. It is particularly for outdoor laborers and vulnerable group such as the elderly and those with chronic illnesses. The phenomenon which indicates a 2.7 to 4.0 ℃ increase every century has contributes to numbers of conditions from heat exhaustion to heat stroke. This situation is imposing a significant pressure on healthcare systems and highlighting the necessity for prevention measures. As heat stroke is a significant danger to health, especially during periods of extreme hot weather. It may lead to multiple medical conditions and perhaps life-threatening problems if not appropriately diagnosed and treated. Thus, the goal of this study is to propose a heat stroke risk detection application that able to help detecting heat stroke risk among user. This application is developed with the implementation of a machine learning that able to detect heat stroke based on users’ medical conditions. Through the application, it enables an early heat stroke risk detection and give suggestions on actions need to be taken to prevent severe heat stroke consequences with the implementation of random forest method.
Today, deep learning (DL) has become a prominent trend within machine learning (ML), offering more advanced pattern recognition and image classification capabilities than traditional ML approaches. Among various DL methods, Convolutional Neural Networks (CNNs) are particularly significant. In this study, we developed an automated classifier capable of categorising a large dataset of fashion clothing items using both DL and ML techniques. Initially, we performed classification tasks employing multiple ML algorithms. Subsequently, we introduced a novel CNN model comprising multiple convolutional layers, a max-pooling layer, and a fully connected layer. Finally, we conducted a comparative analysis of different algorithms. Our programming tools of choice included Python, TensorFlow, and Keras, which are widely used in this field.
This study introduces a Deep Image Prior (DIP)-based medical image denoising method that requires no training data, enhanced through the incorporation of an L1-norm fidelity term. The proposed model, referred to as L1-DIP, addresses the limitations of conventional DIP, which typically relies on an MSE loss and suffers from spectral bias and structural over-smoothing. By replacing the pixel-wise MSE loss with an L1 loss, the method improves robustness to outliers and better preserves anatomical discontinuities, making it particularly suitable for low-dose CT and high-resolution MRI reconstruction tasks. The model employs a U-Net equipped with skip connections, optimized per-image using a fixed random input. The L1 fidelity term guides the reconstruction process toward edge-aware solutions without requiring any external training data. Experiments were conducted on a thoracic CT scan and a coronal pelvic-lumbar MRI image, where additive Gaussian noise was synthetically added to simulate varying levels of corruption. Quantitative evaluations demonstrate that L1-DIP consistently outperforms DIP, DeepRED, and SURE-DIP across all noise levels, in terms of both peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). Visual comparisons confirm that the proposed model achieves clearer boundary definition and better preservation of anatomical structure. Convergence curves for PSNR and SSIM also reveal that L1-DIP converges faster than the DIP. These results highlight the effectiveness of integrating an L1-based fidelity term into the DIP framework, offering a training-free, high-fidelity solution for medical image denoising under complex noise conditions.
The healthcare sector in Oman is widely adopting novel technologies to improve operations, and overall delivery of quality care services. Even though this modernization has great advantages, it has led to an increased number of cyberattacks like sophisticated intrusion penetration attacks that steal data and cripple functions. Such attacks are becoming a significant threat to hospitals, the ministry of health, and insurance companies in Oman. These threats have substantial implications, which include data breaches or losses and financial losses. Oman’s healthcare sector has adopted conventional measures to enhance cybersecurity but these methods are vulnerable to advancing threats or attacks, which requires the implementation of innovative solutions with novel or pioneering technologies. Therefore, the purpose of this paper is to investigate the impacts of artificial intelligence technologies on mitigating cyberattacks or threats in Oman’s healthcare sector. A quantitative method was used to achieve the objectives and research questions. Statistical tools used for analysis include inferential statistics, Chi Square tests and correlation analysis. The findings revealed that artificial intelligence technologies have a greater capacity to conduct continuous evaluation of data breaches, anomalies or behavioral patterns, allowing instant detection as well as responsive action to these attacks. These technologies include predictive analysis methods, algorithms, such as artificial intelligence multisource transfer learning algorithm, neural networks, machine learning, quantum computing, quantum resistant encryption, as well as adaptive cyberthreats learning and pattern recognition. These technologies will also prevent the loss of medical data, decrease threats or adversarial attacks, and enhance cybersecurity in Oman’s healthcare.
Heritage tourism planning normally experiences difficulties in achieving universal accessibility, especially in older cities where cultural monuments are situated within limited and irregular spatial layouts. These environments necessitate inclusive planning practices that take into account the physical, sensory, and cognitive obstacles encountered by various user groups. This research explored the application of Analytic Hierarchy Process (AHP) and Fuzzy Analytic Hierarchy Process (Fuzzy-AHP) to priorities accessibility requirements for determining a pedestrian route in heritage tourism regions. Based on MS1184:2014 and the UNCRPD, a hierarchical framework was established which included three main criteria of mobility and navigation, comforts, and accessibility features, as well as seven sub criteria. Both decision-making models were constructed and analyzed using a simulated pairwise comparison data. The results suggest that, while AHP provides a logical consistency and simplicity, Fuzzy-AHP has a stronger ability to express ambiguity and complexity in expert judgement which giving a larger weight to moderately valued qualities. Although based on simulated data, the study demonstrates the importance of fuzzy-based decision tools in capturing stakeholder variability and promoting inclusive planning. The proposed framework provides a methodological foundation for future empirical research and decision-support systems focused on achieving equitable and sustainable accessibility in heritage tourism settings.
Accurate diagnosis through MRI is the main non-invasive method in the diagnosis of PCas through manual segmentation of the prostate lesion. This process is time-consuming and prone to interreader variability. Deep learning-based image processing has emerged as a solution to the problem. However, most deep-learning-based approaches in prostate lesion segmentation use a one-step direct lesion segmentation approach, which may cause the model to learn from unnecessary inputs that could contribute to noise. To address this problem, some researchers suggested introducing prior prostate anatomical knowledge using a two-step approach that involves segmenting the prostate gland and using the extracted prostate gland as the input for the lesion segmentation model. Despite demonstrating high performance in lesion segmentation, there remains a lack of systematic comparative studies that quantify its practical advantages using standardised metrics, consistent architectures, and the same dataset. In this study, a comparative evaluation of one-step and two-step approaches in prostate lesion segmentation was conducted using the benchmark dataset Prostate158 with DeepLabV3, U-net, TransUNet, and TransAttUnet. The findings reveal that all four models with the two-step approach achieve significantly higher results relative to the one-step approach, with a percentage increase of 160.0
The manual annotation of audio events is a time-intensive and complex bottleneck that severely limits the scalable creation of synchronized content and restricts user personalization in applications. This paper details an automated methodology to address this problem by generating a structured event map from any given MP3 file. The system employs a real-time audio analysis process using the Fast Fourier Transform (FFT) and calculates Spectral Flux to identify significant transient events. A dynamic thresholding technique distinguishes these events, while Beats Per Minute (BPM) analysis refines the output’s temporal accuracy. Experimental evaluation revealed a crucial discrepancy: while the system achieves high temporal precision with a Mean Absolute Error (MAE) ranging from 0.0631 to 0.0991 s, its overall BPM detection accuracy was only 50
The rapid growth of e-commerce platforms has led to massive consumer behavior data, offering opportunities for accurate purchase prediction and personalized marketing. This paper presents a novel hybrid deep learning model that integrates Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks with advanced attention mechanisms to capture complex spatio-temporal patterns in consumer actions. The model employs sophisticated feature fusion and multi-scale temporal modeling techniques to handle heterogeneous data effectively, while ensuring interpretability through tailored attention modules. We validate our approach using the UCI Online Retail dataset, containing over 540,000 transactions. The proposed architecture achieves a prediction accuracy of 78.3
The increasing complexity of spatial decision-making tasks necessitates robust analytical models capable of addressing uncertainty, subjectivity, and large datasets. This study proposes an integrated Fuzzy Analytical Hierarchy Process–Artificial Neural Network (FAHP-ANN) model to enhance the interpretability and computational capacity of Multi-Criteria Decision Analysis (MCDA). The FAHP component captures human judgment through triangular fuzzy numbers to reflect ambiguity in expert assessments, while the ANN component leverages these weights to learn complex patterns across criteria and alternatives. A structured methodological framework is developed, incorporating preference aggregation, ReLU activation functions, Net Flow Score computation, and threshold-based classification for decision support. Monte Carlo simulation and sensitivity analysis are employed to evaluate model stability and robustness. Results demonstrate that the FAHP-ANN model significantly outperforms conventional ANN models in both accuracy (reduction in prediction error) and consistency, particularly under fuzzy environments. The proposed model presents a scalable and interpretable decision-support tool suitable for applications in spatial planning, infrastructure prioritization, and policy evaluation, effectively bridging the gap between human-centric reasoning and data-driven analytics.
Breast cancer histopathological diagnosis is vital for patient prognosis, yet existing automated approaches struggle to reconcile local cellular detail with global tissue context across varying magnifications. To bridge this gap, we present a dual-branch framework that couples a CBAM-enhanced ResNet101 for fine-grained feature recalibration with a hierarchical Swin Transformer that models long-range dependencies. The two branches are aligned through a lightweight multi-resolution fusion module that unifies features from 40 ×, 100 ×, 200 ×, and 400 × scans into a complementary representation. Evaluated on the BreakHis dataset, the proposed model achieves up to 97.8
Fraud detection in online financial transactions has become increasingly important with the rapid growth of digital payment systems and the evolving nature of fraudulent behavior. This literature review examines recent studies on fraud detection, with a specific focus on deep learning and hybrid approaches that combine multiple algorithms to enhance detection accuracy. It examines peer-reviewed works published from 2021 to 2025, emphasizing key trends in the use of deep learning models and their integration with optimization techniques. Frequently used datasets, evaluation metrics, and model architectures are reviewed to provide a comprehensive overview of current practices. In contrast to existing reviews, this study introduces a comparative analysis of several hybrid models, assessing their architectural components, strategies for handling class imbalance, and sequential learning capabilities. Additionally, it offers a critical evaluation of emerging innovations such as attention mechanisms, Generative Adversarial Networks (GANs), and reinforcement learning, highlighting their effectiveness and limitations. The review also identifies core challenges in the field, including the continuous evolution of fraudulent tactics, limitations in available datasets, data imbalance, and the interpretability of models. Finally, the paper discusses potential future directions and emerging solutions that aim to address these gaps. This study seeks to guide researchers and practitioners toward the development of more effective, scalable, and intelligent fraud detection systems by leveraging deep learning and hybrid techniques.
Despite the rapid advancement of financial technologies, the adoption of blockchain in Waqf institutions remains low due to trust issues, perceived risks, and limited digital literacy. This study aims to develop and test a predictive model for individuals’ intentions to use digital Waqf platforms by extending the UTAUT framework. It addresses the need to understand donor adoption by incorporating relative advantage as a mediating factor and knowledge as a moderating variable. The novelty lies in adapting the UTAUT model to the context of Waqf digitalization, offering practical insights for State Islamic Religious Councils (SIRCs) and policymakers to design more effective digital platforms that enhance donor trust and participation. A quantitative method was employed through a survey of 230 respondents, with data analyzed using partial least squares structural equation modeling (PLS-SEM). The findings suggest that performance expectancy and facilitating conditions notably influence digital Waqf adoption, while effort expectancy and social influence do not. Relative advantage mediates these relationships, and knowledge further amplifies the effects. To sum up, the digitalization of Waqf enhances transparency, operational efficiency, and donor confidence, with user support and technological familiarity emerging as key drivers for successful adoption.
This study the usage of sentiment analysis to enhance food tourism in Malaysia by utilizing natural language processing and machine learning. To address data imbalance and variability, a dataset of 360,000 food reviews was pre-processed using methods like text cleaning, tokenization, and resampling. Textual data was converted into machine learning inputs using feature extraction techniques like CountVectorizer and TF-IDF, and sentiments were categorized as either positive or negative. Cross-validation and hyperparameter tuning were used to assess four supervised algorithms including Support Vector Machines (SVM), Logistic Regression, Random Forest, and Multinomial Naive Bayes. With an accuracy of 84.15
The challenge in interpreting data and understanding the message that needed to be conveyed is a significant issue among users. Several studies state that visualizing data are beneficial and lack of understanding of the data can lead to misinterpretations. People can see the data patterns and gain new insights which can be as quick as 300 ms if the data are visualized effectively. Currently, the method used to manage the data of the children’s assessment records is by using a written-form report card. Therefore, this research aim is to implement the Graphical Visualization Technique in presenting student progress data. Another objective is to compare between the usage of current method used in presenting the student progress data and the Graphical Visualization Technique from users’ perspective. There is also an urgent need for a monitoring tool for assisting users’ decision-making process in determining the student progress. This research is significant in order to present the student progress data effectively and also to ensure that the decision-making process to be easier and faster compared to using the current method.
There is a severe lack of effective technology for detecting floods at the community level, which this study aims to address. Residents are frequently unprepared for the sudden increase in floodwater levels. There is an immediate need for an easily accessible and dependable flood monitoring solution due to the significant risk to public safety and property posed by the lack of early local warning capabilities. The research employs IoT devices, including ultrasonic sensors, water level sensors, and DHT22 (Digital Humidity, and Temperature Sensor), strategically positioned to monitor water levels in real-time. This data is utilized in machine learning models such as decision trees, random forests, and linear regressions to predict floods. The IoT-based system employs real-time monitoring and a machine learning algorithm to provide early flood warnings. Among all models utilized in this research, the Decision Tree achieved the highest accuracy at 98.14
