Applied Mathematics and Computational Intelligence (AMCI)

Clustering Shariah-Compliant Stocks In The Construction Sector On Bursa Malaysia

2025-03-27T15:31:05+00:00

The complexity and volatility of stocks have made it difficult for investors to screen ethical and profitable options. Finding the ideal portfolio requires navigation through market uncertainties, fluctuations and fundamentals. Hence, this paper explores strategies to develop a risk-category portfolio to suit investors while adhering to Shariah guidelines. The main goals are to determine distinct clusters of Shariah-compliant stocks for construction companies in Bursa Malaysia and to categorize the cluster of stocks using selected stock indicators. K-Means Cluster analysis was employed to segment the construction stocks. Data on indicators were assessed from a finance portal from 2022 to 2024. The findings revealed three different clusters. Cluster 1 has the highest stock price and market capitalization, making it the most stable and lucrative option for long-term investors. Cluster 2, which exhibited moderate risk investment, is ideal for investors prioritizing capital preservation. Cluster 3, with 56 stocks, displayed high volatility and lower prices, appealing to risk-tolerant investors seeking growth opportunities. Based on these findings, the study aims to support investors in making informed, ethical, and strategically sound decisions in line with Shariah principles.

TPOT-MLP-SVM: Hybrid Model of Multilayer Perceptron with Support Vector Machines Based on Genetic Programming for Predictive Analysis

2024-12-02T02:42:17+00:00

Nowadays, the integration of machine learning has drawn significant attention due to its effectiveness and robustness. Automated machine learning (AutoML) has transformed the field of artificial intelligence by developing an effective model to solve predictive tasks. Despite the performance of the conventional model, there is a need for improvement to achieve a better and more effective model. In this paper, we proposed TPOT-MLP-SVM, a novel hybrid model to enhance model performance that leverages genetic programming (GP) by integrating multilayer perceptron and support vector machine to improve predictive scores and robustness. The proposed method utilizes GP to automatically search and optimize the structure and parameters of both MLP and SVM variables, thereby minimizing manual input and maximizing predictive performance. Lastly, a real dataset was used to train, test, and validate the proposed model using Python software. The performance was evaluated based on accuracy, precision, sensitivity, specificity, f1-score, and ROC-AUC. Experimental results on benchmark datasets demonstrate the effectiveness and better performance of TPOT-MLP-SVM over stand-alone MLP and SVM models and other hybrid techniques. Overall, TPOT-MLP-SVM is a potential tool for predictive analysis, as it integrates MLP and SVM in a single system guided by genetic programming for early medical diagnosis in healthcare.

Detection of Diabetic Retinopathy Using a Transfer Learning Approach with DarkNet19

2024-12-11T05:52:11+00:00

The early detection of diabetic retinopathy (DR) in fundus images is crucial for preventing vision loss in diabetic patients. Deep learning models have played a significant role in advancing DR detection. This paper explores the relatively unexplored DarkNet19 model’s performance in comparison to well-established models like ResNet18 and ResNet50 for DR detection. A balanced dataset of healthy and DR images was created through standardization and augmentation techniques. The models underwent binary classification training and testing, and their performance was evaluated using accuracy and precision metrics. DarkNet19 outperformed the other models, achieving higher accuracy (0.7347) and precision (0.9233), demonstrating its potential to enhance early DR diagnosis and reduce the risk of vision loss. This research contributes to the field of DR detection, highlighting the effectiveness of DarkNet19.

Outlier Map of Classical and Robust Principal Component Analysis

2025-04-27T05:18:36+00:00

Classical Principal Component Analysis (CPCA) is widely used for dimensionality reduction, but it is highly sensitive to the presence of outliers, leading to distorted covariance estimates and unreliable principal components. To address this, Robust PCA (ROBPCA) integrates robust covariance estimation and projection pursuit to minimize the effects of outliers. Although CPCA and ROBPCA are often utilized for high-dimensional data , it is equally effective in low-dimensional settings, particularly when handling large datasets. This research illustrates the benefit of ROBPCA over CPCA by analyzing a large-scale gene expression dataset with 22 features and 47231 observations to demonstrate its efficiency in identifying and classifying outliers using outlier maps. Findings reveal that CPCA misidentifies outliers, leading to inflated variance structures and poor principal component estimation, whereas ROBPCA successfully isolates outliers, preserving data integrity and enhancing interpretability. This research emphasizes how ROBPCA improves data reliability and offers a reliable method for identifying outliers.

Trends and Patterns in Financial and Retirement Planning Research: A Scientometric Perspective

2025-03-25T15:20:53+00:00

Financial planning has emerged as a global concern encompassing not only the management of funds but also encompassing the projection of future income, asset values, and expenditure schedules. This study utilized bibliometric analysis to uncover underlying publication patterns related to financial and retirement plans. It aimed to provide a foundation for a deeper understanding of historical, current, and future trends in these fields. Research articles spanning from 1954 to 2023 and 1976 to 2023 were sourced from the Scopus and Web of Sciences (WOS) databases respectively, totaling 1985 and 1333 articles. The analysis revealed the United States as a leading contributor in publishing research articles on financial and retirement plan, along with the most productive institutions and authors. The USA also emerged as the most active collaborator internationally, engaging with 74 countries. Research hotspots were identified via keyword co-occurrence analysis, highlighting "retirement planning" and "retirement" as prominent areas. This analysis emphasized the critical role of financial and retirement plans and identified several research gaps in their practical application, suggesting potential areas for future research. It also underscores the active involvement of Malaysian universities in the development of this field and reveals critical research gap in its practical application. Future research endeavors should prioritize promoting the utilization of financial and retirement plans. Nonetheless, this bibliometric study holds the potential to serve as a foundational step in supporting future meta-analyses and structured literature reviews.

A Combination of Power and Exponential Series for the Development of Hybrid Block Integrators for the Solution of First Order Ordinary Differential Equations

2024-12-12T07:05:52+00:00

The research article, presented the derivation and implementation of some block integrators at step sizes and respectively for the solution of first order ordinary differential equation via power and exponential series as basic functions. The derived methods were found to be zero-stable, consistent and convergent. Also for the values of present in the basis were tested with various numbers within the number system, the schemes obtained remain the same. The newly derived methods displayed their superiority when tested on some real life problems and some non-linear differential equations.

Modelling Economic Growth: Panel Data Approach

2025-03-22T19:00:18+00:00

The Gross Domestic Product (GDP) plays a pivotal role as a key economic indicator. There are numerous factors that contribute to the formation of GDP. Hence, this research attempts to identify the best-fit panel model for GDP among selected countries in Southeast Asia and determine the significant factors influencing their GDP. The data used is the panel data that consist of GDP, Consumer Price Index (CPI), Unemployment Rate (UR) and population growth (POP) from 2003 to 2022 for Malaysia, Thailand, Indonesia, Brunei and Singapore. The methods of pooled, fixed and random effects models are employed. The fixed effects model reveals substantial impacts of variables like UR, POP, and CPI on GDP. The random effects model, validated through the Breusch-Pagan test, demonstrates superior adaptability to country heterogeneity. The Hausman test supports the random effects model as a more reliable framework than fixed effects. The unemployment rate and population growth affect significantly towards GDP.

Assessing Hierarchical Component Model for Work-Life Balance Strategies and Entrepreneurial Competencies: Reflective-Reflective Model

2025-03-15T19:02:42+00:00

A hierarchical component is a model with complex systems in a hierarchy where higher-level components depend on a composed of lower-level components. Each level represents a distinct level of abstraction or complexity. The objectives of this study aim to assess the measurement and structural assessment of hierarchical component model (HCM) of Work-Family Enrichment (WFE) and Family-Work Enrichment (FWE) on entrepreneurial competencies among women entrepreneurs in Malaysia. The model applies a reflective-reflective approach, using a quantitative method with survey data collected from 284 women entrepreneurs in Malaysia, and assessed using the Structural Equation Modeling (SEM). The analysis and result shows the evidence that the items and constructs in this model were passed the assessment of first and second order of measurement and structural model. These results validated the hierarchical component model for these constructs, ensuring the dimensions and indicators are reliable and valid for understanding the relationships between WFE and FWE. The implications of this study extend to research on hierarchical component models, demonstrating their utility in exploring multi-dimensional constructs by providing a robust framework for future investigations into work-life balance strategies.

Hemaclassify: Web-based Blood Diseases Classification System Utilizing High-Resolution Neural Networks

2025-01-06T01:18:06+00:00

Blood classification plays a pivotal role in diagnosing various blood-related diseases, including Acute Lymphoblastic Leukemia (ALL), Acute Promyelocytic Leukemia (APML), Chronic Myeloid Leukemia (CML), Iron Deficiency Anemia (IDA), Thalassemia Major, Thalassemia Minor, and Normal Blood Cells. The traditional approach to blood classification involves manual analysis by hematologists. This intensive work process takes time and sometimes requires immediate diagnosis for cases that can be fatal. Therefore, to streamline this process, deep learning and image processing techniques will be utilized. This project aims to address these challenges by developing a web-based expert system for the automatic classification of blood samples using High-Resolution Neural Networks (HRNet). Inspired by the HRNet architecture originally designed for computer vision tasks, the system maintains high-resolution representations throughout the classification process. An innovative approach connects high-to-low resolution convolution streams in parallel, facilitating the preservation of high-resolution information. The exchange of information across resolutions enhances the semantic richness and spatial precision of the resulting representations. The system is capable of classifying blood samples into seven distinct categories as mentioned earlier. The proposed web application streamlines the diagnostic process, offering a faster and more accurate classification of blood samples. By automating this critical task, it supports hematologists in making more efficient diagnoses and helps prevent the progression of conditions like CML into more dangerous and acute phases. The dataset used is divided into a ratio of 8:1:1, representing the training, validation, and testing datasets. The total number of images used for training, validation, and testing are 549, 66, and 75, respectively. Thus, the total image dataset used in this project is 690. The system is targeted to achieve an accuracy rate of 99.89\%, demonstrating its effectiveness in blood classification. The HemaClassify system utilizes a REST architecture, with the frontend developed using NextJS, TailwindCSS, and ShadCN UI, and the backend implemented with a Flask API and the HRNet model. MySQL is used for database management. The user interface (UI) has been evaluated through User Acceptance Testing (UAT) conducted by Medical Laboratory Technologists (MLT) and hematologists from Hospital UKM, ensuring the system meets the requirements and expectations of end-users. The UAT results indicate a high level of user satisfaction, with most users rating the system 4/5 or 5/5 for its usability and functionality. This project represents a significant advancement in the field of blood classification, contributing to more timely and precise diagnoses, which are essential for effective disease management and treatment.

A Review of Convolutional Neural Network-Based Automatic Lane Detection Methods on the TuSimple Dataset

2025-02-25T02:57:46+00:00

This review explores and compares various automatic lane detection methods, shedding light on their strengths, weaknesses, and advancements. The study analyzes a diverse range of techniques, including model-based and deep learning-based approaches, employed in road lane detection. The review highlights the advantages and disadvantages of each method, providing a nuanced understanding of their performance metrics, accuracy, and applicability under different scenarios. It dives into the evolution of lane detection algorithms, emphasizing recent breakthroughs in the field. The comparison section systematically evaluates the effectiveness of these methods, considering factors such as computational efficiency, robustness in challenging conditions, and adaptability to diverse environments. It aims to guide researchers, practitioners, and developers in choosing suitable lane detection methods based on specific use cases and requirements. Ultimately, this review contributes to the ongoing discourse in the area of autonomous driving, intelligent transportation systems, and computer vision, offering valuable insights for the continuous improvement of automatic lane detection technologies.