Asian Journal of Research in Computer Science
Not a member yet
792 research outputs found
Sort by
Deep Learning-Based Skin Type Identification and Personalized Skincare Recommendations
No full textThe skin is the outermost layer of the body and is crucial for women\u27s beauty and health. Understanding one\u27s skin type is essential for effective skincare, as a lack of knowledge can lead to adverse effects. This study proposes a deep learning-based system designed to identify women\u27s facial skin types and provide personalized skincare recommendations. Leveraging the MobileNetV2 architecture as the base model, a convolutional neural network, the model was trained to classify five skin types: Dry, Oily, Normal, Combination, and Sensitive. Data were collected from cosmetology specialists and online datasets, with preprocessing and image augmentation to enhance the dataset. The approach involved modifying the MobileNetV2 by adding GlobalAveragePooling2D and custom dense layers tailored for skin type classification. The developed model achieved a testing accuracy of 93.78% and was integrated into a Flask-based web application. This application allows users to upload facial images, provide their age and allergy status, and receive tailored skincare routines based on the predicted skin type. User testing demonstrated a high level of satisfaction, with 64% of respondents finding the skin type predictions accurate and 61% considering the skincare recommendations useful. The website\u27s interface received positive feedback, with 95% of respondents rating it as user-friendly. While the model performed well in classifying most skin types, it encountered difficulties in differentiating between visually similar types, such as Combination and Sensitive, where misclassifications occurred due to overlapping visual features. Additionally, the study faced limitations stemming from image quality variability, and the constraints of rule-based recommendations. Furthermore, the current dataset did not fully represent all demographic groups and diverse skin tones. Overall, this study contributes to the scientific community by presenting a practical, real-time application of AI in dermatology that bridges the gap between technology and everyday skincare needs
A Geometric Perspective on Some Schroder Sequence Spaces
No full textInvestigation of some structral properties of some sequence spaces constructed using Schr¨oder numbers has recently become important. Some of the fundamental works on the investigation of the properties of these sequence spaces are given in (Daglı, 2022a,b) and (Yilmaz et al., 2025). In this work we are going to investigate some geometric properties such as rotundity and uniformly smoothness of the sequence spaces λ (S) where λ is c0, ℓ∞ or ℓp and 1 ≤ p < ∞
From Theory to Application: Evaluating the Efficiency, Scalability and Predictability of Classical and Modern Sorting Algorithms in Real-Time Systems
No full textEfficient and predictable sorting is critical for real-time computing applications, including embedded systems, financial trading platforms, and sensor networks. In this paper, a strict comparative analysis of five popular sorting algorithms, namely Heap Sort, Merge Sort, Quick Sort, Tim Sort, and Intro Sort, in different input situations, i.e., almost sorted, random, and reverse-sorted data, is provided. Every algorithm was tested using 30 independent trials per condition so that the statistical strength is attained. Findings have shown that Tim Sort has a consistently better performance, the lowest execution time (6 ms) and the smallest memory footprint (7.80 MB), and thus its flexibility to data patterns of various types. Quick Sort is fast in cases that are average-case but has a higher memory overhead in cases where the input is reversed sorting. Hack Sort is stable with random data and Quick Sort is fast with average-case data. Merge Sort is predictable but resource-intensive, as the execution time goes up to 28 ms, and memory consumption is up to 8.34 MB. The performance of Intro Sort is balanced and has execution time of 11-13 ms and a memory size of less than 7.90 MB, regardless of the dataset. The significance of these differences can be proved by statistical procedures and correlations such as one-way ANOVA and post-hoc tests (Tukey) with the significance level of p less than 0.001. The combination of empirical assessment with strong statistical verification makes this study offer practical advice on the choice of algorithms in the latency sensitive computing environment. Further studies will elaborate this framework to parallel and distributed versions, which allows scalable and high performance sorting in the present-day real-time systems
A Cloud Implementation Assessment System for SMEs
No full textThe emerging impact of cloud computing in businesses has become a great driving force for business success and cannot be understated. The lack of a tailored functional application software, even though a framework exists, poses a significant challenge to SMEs. Without a digital corresponding software solution, SMES struggle to conduct a practical assessment since the framework offers only theoretical guidance. A well-developed cloud implementation assessment tool is essential for SMEs to thrive.
The main aim of the study was to develop a robust cloud implementation assessment system for SMEs to evaluate their cloud implementation readiness. Design Science Research Methodology (DSRM), is a kind of research methodology which aims at developing and evaluating innovative artefacts, in the form of models, frameworks, designs, or prototypes, to solve complex and practical problems. Framework Usability Validation used. On the Clarity of the framework, none of the respondents strongly disagreed, representing 0%. 8 respondents, representing 15.4% disagreed, and 3 respondents, representing 5.8% were unsure. The system was developed using the Django Python web framework. Out of 43 respondents, 18 participants (42%) found the Learn Cloud Phase helpful, 3 participants (7%) found the Assessment and Cloud Choice Phase helpful, and 15 participants (35%) found the Migration Phase helpful. Additionally, 4 participants (9%) reported that all phases were helpful, 2 participants (5%) indicated that none of the phases were helpful, and 1 participant (2%) was uncertain about the helpfulness of any phase.
In this paper, the design science Research (DSR) method was utilised to develop and evaluate the cloud implementation assessment system for SMEs. Key features of the developed system include learn cloud, assessment, cloud choice and migration phase. The developed system was said to be very useful and suitable, solidifying its value as a helpful resource for small businesses aiming to utilise cloud technology. A cloud assessment system was developed to aid in assessing SMEs\u27 cloud readiness and implementation. Generally, SMEs responded positively to the cloud computing implementation framework
Comparing the Performance of Convolutional Neural Networks and Vision Transformers in Object Detection: A Review
No full textThis review studies the evolution of object detection methodologies, from traditional to modern deep learning techniques, including CNNs (Convolutional Neural Networks), YOLO (You Only Look Once) variants (v1–v8), and ViTs (Vision Transformers). A systematic analysis of 49 studies shows that CNNs are robust on small datasets and in real-time applications. In contrast, ViTs excel at handling complex relationships and adversarial conditions due to their self-attention mechanisms. Hybrid models combining CNNs and ViTs show promise for improved accuracy and efficiency but usually require further validation. Key challenges include computational demands, dataset diversity, and generalisation across domains. Despite significant progress, there is limited consolidated analysis comparing CNNs, YOLO, and ViTs across diverse datasets and real-world constraints. The comparison in this study may level the ground for researchers to explore new gaps in the future, and results not only in reinforcing the potential of object detection techniques but also provide useful insights for researchers and practitioners aiming to balance performance with computational cost in real-world detection scenarios. Future research should prioritise hybrid architectures, edge deployment, and standardised benchmarking to advance object detection in different domains such as surveillance, healthcare, quality control, inventory management, and autonomous systems.
 
A Survey on Binary Tree-Based Approaches for Data Transmission in Mobile Ad Hoc Networks
No full textA thorough analysis of the current binary tree-based data distribution techniques in MANETs is the goal of this paper. MANET communication is highly dynamic, necessitating effective data transmission methods to improve network stability while also saving energy. Binary tree topologies work in tandem with routing and data aggregation to improve scalability, reduce latency, and increase energy economy. The paper investigates several binary tree algorithms that are appropriate for data and security structures, as well as routing techniques. Similar to previous MANETs, the network has three main issues: security threats, energy constraints, and mobility issues. Key features of modern algorithms are briefly discussed in the study, along with the benefits and drawbacks of tree-based systems. The study itself outlines the goals and directions for further investigation just to find the best network throughputs while staying within the constraints of dynamism\u27s limited computing and energy resources
Integrating Newcomers: Effective Models for Organizational Adoption
No full textThis study utilizes a comparative approach to explore and contrast various models for newcomer integration within organizational contexts. By conducting a systematic literature review, the research identifies key components, outcomes, and challenges associated with each model. Through a detailed analysis of models such as the Developers Joining Model, Onion Model, Identity Socialization Model, Four C\u27s Model, and Traditional On-boarding Process, the study evaluates their performance across critical factors including onboarding plans, mentorship, feedback, cultural fit, flexibility, and role clarity.
The findings reveal distinct strengths and limitations for each model, highlighting their varied effectiveness in promoting employee engagement and retention. The Identity Socialization Model and Four C\u27s Model, for instance, excel in fostering long-term engagement, while the Traditional Model supports initial integration but may limit personal identity expression. The study also addresses potential drawbacks, such as resource intensity or context-specificity, proposing mitigation strategies such as phased implementation and role development.
Ultimately, this research provides actionable insights and practical recommendations for HR leaders aiming to enhance their newcomer integration strategies. By aligning the on-boarding strategies with organizational goals, organizations can foster both the immediate and long-term benefits to enhance the overall satisfaction and retention of employees
CTI Integration in Contact Centers: A Comparative Analysis of Security, Scalability, and Challenges in Legacy vs. Cloud-Based Systems
No full textAIM: Scope of this work aims to explore the integration of Computer Telephony Integration (CTI) in both legacy and cloud-based contact center systems, examining security challenges and integration methods for achieving robust Computer Telephony Integration implementations.
Study Design: This study provides a comparative analysis of techniques, methodologies, and integration challenges associated with CTI in legacy on-premises systems and modern cloud-based solutions, with a particular focus on security considerations in both contexts.
Place and Duration of Study: This study is based on a review of contact centers in retail versus mid-sized tech companies, focusing on solutions implemented between 2018 and 2024.
Methodology: This study compares Computer Telephony Integration in both legacy and cloud-based contact center systems, focusing on integration processes, security concerns, scalability, challenges, and feature differences. It uses a combination of key evaluation criteria (covering core components, flexibility, security practices, and scalability), case studies, financial implications, and a feature comparison matrix to explore CTI integration in traditional on-premises environments versus modern cloud-based solutions. Security has been the primary concern, and this included encryption, authentication and compliance with regulations such as the General Data Protection Regulation (GDPR) and the Health Insurance Portability and Accountability Act (HIPAA).
Results: The study found that cloud-based contact centers offer more flexible and scalable integration options, leveraging modern APIs, cloud-native services, and omnichannel support. Legacy systems, while reliable, face challenges in scalability and integration, requiring complex middleware and custom solutions. Security is a significant concern for both environments; however, cloud solutions are enhanced by regular updates, improved compliance standards, and centralized management. Data encryption, secure voice protocols, and authentication were identified as the foundation for security for all types of Computer Telephony Integrations in either category of systems. In addition, the study found that cloud-based systems are being accelerated as more flexible and economically viable solutions are being demanded.
A unique contribution of this study is the development of a security-focused comparative framework for this integration in legacy and cloud-based contact centers. By analyzing security practices, scalability challenges, and integration methodologies, this study bridges a critical knowledge gap, offering practitioners a comprehensive guide to CTI decision-making.
Conclusion: Computer Telephony Integration (CTI) plays a critical role in modern contact centers. While legacy systems continue to serve their purpose, cloud-based solutions offer superior flexibility, scalability, and security. Organizations must prioritize robust security practices when implementing them, regardless of the platform. Future research should focus on advancing AI-enhanced security frameworks to address evolving threats, evaluating hybrid CTI models that integrate legacy and cloud components, and testing the impact of real-time analytics on customer satisfaction. Such advancements will ensure data privacy, regulatory compliance, and the continuous improvement of customer engagement and operational efficiency
Dropout: An Effective Approach to Prevent Neural Networks from Overfitting
No full textOverfitting remains a significant challenge in training neural networks, often leading to poor generalization on unseen data. Dropout has emerged as a powerful regularization technique to mitigate overfitting by randomly deactivating neurons during training, thereby preventing co-adaptation of features and encouraging diverse representations. This paper explores the theoretical foundations and practical implementations of dropout across various neural network architectures, including Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). Through empirical analysis on benchmark datasets such as CIFAR-10, MNIST, and others, dropout is shown to improve model robustness and accuracy significantly. The study also compares dropout with alternative regularization methods, such as weight constraints and batch normalization, highlighting its effectiveness in diverse scenarios. Despite its success, dropout\u27s performance is influenced by hyperparameter tuning and dataset characteristics. The paper concludes by discussing limitations, such as computational overhead, and proposes directions for optimizing dropout for specific applications, including dynamic dropout rates and hybrid regularization techniques
Post-Quantum Cryptography and Encryption Standards: Safeguarding Patient Data against Emerging Cyber Threats in Telemedicine
No full textPost-quantum cryptography (PQC) is a critical innovation for securing telemedicine patient data against emerging quantum computing threats. This study evaluates the vulnerabilities of existing encryption frameworks using data from the U.S. Department of Health & Human Services (HHS) Healthcare Data Breach Report. A weak correlation (r = -0.087) between encryption strength and breach severity suggests that while stronger encryption reduces the number of compromised records, factors such as system misconfigurations, phishing attacks, and insider threats remain significant contributors to data breaches. The study further benchmarks the performance of four NIST-approved PQC algorithms—Kyber, Dilithium, Falcon, and SPHINCS+—by analyzing encryption time, decryption time, key size, computational overhead, and storage requirements. Benchmarking data from the National Institute of Standards and Technology (NIST) is statistically evaluated using one-way ANOVA, which identifies significant performance differences among the PQC algorithms (p < 0.05). Falcon demonstrates the highest efficiency, with an encryption time of 17.16 ms, a decryption time of 18.59 ms, and optimized storage (2.05 MB), making it well-suited for real-time telemedicine applications. Institutional readiness for PQC adoption is assessed using the Healthcare Information and Management Systems Society (HIMSS) Cybersecurity Survey, identifying technical expertise (6.97/10) and infrastructure readiness (6.97/10) as the strongest adoption determinants. Based on the findings, this study recommends prioritizing Falcon for PQC adoption in telemedicine due to its superior efficiency, enhancing encryption key management protocols to mitigate insider threats, and strengthening cybersecurity infrastructure to address encryption misconfigurations. These measures will ensure that telemedicine systems remain secure, resilient, and capable of mitigating quantum-era cyber threats