University of Ibadan Journal of Science and Logics in ICT Research

Development of an Improved Mayfly Algorithm Based Convolutional Neural Network for Pulmonary Diseases Recognition System

2025-12-26T07:47:25+00:00

Pulmonary diseases impact the respiratory system. Convolutional Neural Network (CNN) is used for detection and recognition of pulmonary diseases; however, it suffers from hyperparameter selection and overfitting problems. Existing optimization techniques such as the Mayfly Algorithm (MA) also require initial parameter tuning and exhibit slow convergence behaviour. This research developed a Roulette Chaotic Mayfly Algorithm (RCMA) based on CNN (RCMA-CNN) for pulmonary diseases recognition. X-ray images including normal and pulmonary diseases cases were obtained from Kaggle and pre-processed for the desired image quality. The RCMA was formulated using Roulette wheel selection to model attraction deterministically and Chaotic Sinusoidal Map Function to balance exploration and exploitation in the MA. RCMA was applied to optimize CNN hyperparameters including number of layers and batch size at the convolutional layer. This was implemented in MATLAB (R2020a) and compared with MA-CNN and CNN in terms of false positive rate, sensitivity, specificity, accuracy and recognition time. At optimal threshold of 0.75, RCMA-CNN gave false positive rate of 1.43%, sensitivity of 98.06%, specificity of 98.57%, and accuracy of 98.32%. RCMA-CNN recorded a recognition time of 76.81 seconds, which was better than that of MA-CNN and CNN. The RCMA CNN model significantly outperformed both MA-CNN and standard CNN.

Development of Comparative Fake Transactions Alert Detection Models Using Machine Learning

2025-12-26T06:51:46+00:00

Fraudulent payment evidences are the current techniques criminals employ to defraud businesses and small scale enterprises. Researchers have processed transactions textual data however, there is still the challenge in the ability to differentiate between legitimate and fake transaction alerts. In Nigeria, fake transaction alerts pose significant challenges for financial institutions and individuals losing on their hard earned assets, citizens are sceptical on electronic transactions and several Point of Sale (PoS) businesses have fallen victims. Hence, this study was aimed at the development of a better fake transaction alert detection model to distinguish fake transaction alerts. Artificial Commercial Data for Fraudulence Discovery was collected from Kaggle website. The collected data was pre-processed. Data imbalances were handled. Support Vector Machine (SVM) and Random Forest (RF) algorithms were ensembled to simulate fake transactions alert detection models using MATLAB programming. They were trained and tested with 70% training and 30% testing datasets, respectively. Performance evaluation was done on RF and SVM classifiers using exactness, precision, recollection, F-measure as benchmarks. The data record employed for this study had 1,048,575 transactions alerts. At performance evaluation, RF model had exactness, precision, recollection and F-measure values, 97.6, 97.48, 97.54 and 97.51%, respectively. Its RMSE was 0.02376. Moreover, SVM model had exactness, precision, recollection and F-measure values of 96.1, 96.88, 98.38, and 97.63%, respectively, it has RMSE of 0.03911. Random Forest algorithm was more suitable for the development of the fake transactions alert detection because it had higher performance. This model could be adopted by financial related institutions.

Design and Implementation of Monopoly Game using Yoruba Language

2025-12-23T19:37:46+00:00

Monopoly is a classic board game that simulates real estate trading and economic principles. Played by two to eight players, its objective is to accumulate wealth through strategic property acquisition and development, while bankrupting opponents. The game is known for its iconic board layout featuring properties named after real streets and locations, alongside utilities and railroads. Players roll dice to move around the board, purchasing properties they land on and collecting rent from opponents who land on their owned properties. Monopoly also incorporates chance and community chest cards that can either benefit or hinder players' financial positions but all the features use in playing the game was written in English language which is a common language in a large community. However, some non-educated people most especially people from western part of Nigeria always find interest in this game but their inability to read characters written in English has incapacitate them from playing the game. To resolve this problem a monopoly game using Yoruba Language that will allow larger audience and Yoruba like to participate in the game players and to monopolize property groups to increase rental income, making negotiations and trades essential components of game play was developed. The research methodology, flow chart and the board interface system are explored. The flow chart illustrates interactions between player, color representing each player, players turn, money deduction or addition when landed on each corner on the board. The new system, a monopoly game using Yoruba Language, is user-friendly interface and the players were able to play it conveniently with no stress.

Pipeline Leakage Detection and Monitoring Model using Enhanced Multiple Signal Classification Algorithm and Hybrid Acoustic Emission Techniques

2025-12-23T09:23:09+00:00

The consequences of pipeline leakages pose great multifaceted hazards, including carcinogenicity and cytotoxicity in humans exposed to leaked toxic substance from pipelines. Pipeline leak also causes environmental contamination of soil resulting to environmental pollution, fire disaster and even loss of life. Therefore, pipeline leakage detection monitoring is a crucial concern in pipeline industry for ensuring the safe and efficient operations. Background noise and detection of single leak are significant limitations of the existing pipeline monitoring and leakage detection techniques. These undesired noises can arise from multiple sources, including environmental, proximity industries, pipe vibration, and electronic interferences. This study therefore optimizes the conventional Multiple Signal Classification (MUSIC) algorithm and Acoustic Emission (AE) technique with the aim to develop a novel technique to address the effect of the background noise. The proposed method combines the advantages of the MUSIC algorithm and AE techniques with real-time monitoring to promptly and accurately detect leakages in pipeline systems. The model achieved Accuracy of 95.5%, Sensitivity of 75%, Mean Detection Time of 1.02 seconds and Response Time of 1.06 seconds. These quantitative results demonstrate the effectiveness of our proposed Enhanced MUSIC algorithm and Hybrid AE technique (Enhanced-MUSICHAE) to detect and monitor pipeline leakage. This has the potential to improve pipeline safety, reduce economic losses, and minimize environmental damages.

Application of Machine Learning Algorithms in Predicting the Toxicity of Chemical Compounds for Safer Pharmaceuticals

2025-12-23T08:22:41+00:00

The development of safe pharmaceuticals requires accurate and efficient prediction of chemical toxicity to minimize adverse health risks and reduce reliance on costly and ethically challenging animal testing. This study investigates the application of three machine learning (ML) algorithms—Random Forest (RF), Support Vector Machine (SVM), and Linear Regression (LR)—for predicting the toxicity of aromatic chemical compounds. A dataset of 11,001 compounds was curated, preprocessed, and analyzed using molecular descriptors such as molecular weight, lipophilicity, and polar surface area. Model performance was evaluated using accuracy, precision, recall, F1-score, and specificity. Results showed that the Linear Regression model performed poorly, with accuracy around 52%, indicating limited suitability for toxicity classification. The SVM model achieved substantially better results, with an accuracy of 80%, demonstrating its effectiveness in capturing nonlinear structure–toxicity relationships. Notably, the Random Forest model outperformed both, achieving perfect classification accuracy (100%) across all metrics, with zero false positives and false negatives. Feature importance analysis revealed that descriptors such as Topological Polar Surface Area and Molecular Fractional Polar Surface Area were key contributors to toxicity prediction. The findings demonstrate that Random Forest is a robust and interpretable tool for early toxicity screening, offering both predictive accuracy and insight into molecular features driving toxicity. By integrating ML models into pharmaceutical research pipelines, drug discovery can be accelerated, costs reduced, and ethical imperatives met by minimizing animal testing. Future work should focus on external validation, hybrid model development, and explainable AI techniques to enhance generalizability and regulatory acceptance.

Design Cycle Methodology for Developing AI Microservices Frameworks: A Case Study

2025-12-26T09:53:53+00:00

The integration of Artificial Intelligence (AI) and microservices is increasingly recognised as a pathway to building scalable, reusable intelligent systems. Yet much of the existing work remains implementation-driven, with limited methodological grounding, which restricts reproducibility and generalisability. This paper presents a case study applying the Design Cycle Methodology (DCM) to the systematic development of an ontology driven AI microservice framework, the AI Microservice Agent. The study addressed four research questions: whether semantic registration improves service discovery efficiency, how it supports scalability under load, what computational trade-offs are introduced, and how well the approach generalises across domains. A proof of-concept text classification microservice was semantically described using OWL service descriptors and retrieved via SPARQL queries, illustrating the operational role of semantic registration. Comparative experiments against a monolithic system demonstrated a reduction of up to 35% in discovery latency, stable throughput under increasing client requests, and robustness under failure conditions with only minor reasoning overhead. Cross-domain validation with text and image services achieved 100% successful integration, confirming generalisability. To our knowledge, this is the first study to embed AI microservice development within DCM, providing methodological traceability between objectives, design stages, and empirical findings. By releasing ontology files, service descriptors, and containerisation artefacts, the work contributes a reproducible framework that advances discovery efficiency, scalability, and adaptability in AI microservices research.

A Machine Learning Framework for Classifying Haemoglobin Levels in Sickle Cell Anaemia Patients

2025-12-23T05:36:05+00:00

Sickle Cell Anaemia (SCA) significantly impacts haemoglobin (HGB) levels, leading to severe health complications with high mortality rates. In Nigeria, about 2% of newborns, approximately 150,000 annually, are diagnosed with SCA. Accurate HGB monitoring is essential for effective disease management, yet traditional methods are labour-intensive and prone to errors. This necessitates automated and reliable diagnostic techniques like machine learning (ML) for improved SCA management. This study classifies HGB levels in SCA patients using clinical records and ML techniques. A dataset of 364 records (203 female population) was obtained from Kaggle; a public data repository containing eleven (11) features namely: age, sex, red blood cell (RBC) count, packed cell volume (PCV), mean corpuscular volume (MCV), mean corpuscular haemoglobin (MCH), mean corpuscular haemoglobin concentration (MCHC), red cell distribution width (RDW), total leukocyte count (TLC), platelets per cubic millimeter (PLT/mm³), and haemoglobin (HGB). Two ML models, Logistic Regression (LR) and Support Vector Machine (SVM), were used with two feature selection methods: all features and selected features. The latter identified age, RBC, PCV, MCV, and HGB as key predictors. Continuous HGB values were categorized into (1) low, (2) normal, and (3) high using standard medical metrics. SMOTE analysis was also carried out to mitigate class imbalance. SVM with a Radial Basis Function (RBF) kernel achieved 84.90% accuracy and AUC-ROC of 93.40%, while LR underperformed with 79.50% accuracy and AUC-ROC of 90.90%. Using all feature selection, SVM improved to 91.80% accuracy and AUC-ROC of 98.20%, with LR achieving accuracy of 93.20% and AUC-ROC of 98.90%. Both models demonstrated high accuracy, with LR excelling using all features, while SVM performed better with selected features. Future work will involve the use of primary datasets, additional feature selection techniques and ML algorithms, and incorporate the use of Haemoglobin variants to provide further insight into SCA progression and in turn offer personalized treatment.

Practical Applications of Network Management Tools in Emerging Technologies

2025-12-26T08:40:40+00:00

The rapid evolution of emerging technologies—such as the Internet of Things (IoT), edge computing, 5G network slicing, and artificial intelligence (AI)—has significantly reshaped network management practices. As networks become increasingly complex, large-scale, and diverse, traditional approaches relying on manual oversight and static, rule-based systems are no longer sufficient. To address these growing demands, modern network management is shifting toward intelligent, automated solutions capable of real-time analysis, dynamic resource allocation, and improved security. This article examines the practical applications of advanced network management tools, with a particular emphasis on AI-driven monitoring, anomaly detection, automation, and the move toward standardizing network intelligence. Based on recent technological developments from 2019 to 2025, it evaluates how these tools contribute to building more resilient, adaptive, and secure networks. The discussion highlights key advantages, including predictive maintenance, faster fault detection, optimized traffic handling, and proactive threat response. However, it also addresses limitations such as privacy risks, potential algorithmic bias, and integration challenges with legacy systems. Emerging trends such as self-healing networks, federated learning, and intent-based networking are explored as future directions for scalable and intelligent infrastructure. By addressing both the benefits and challenges, this article emphasizes the essential role of AI enhanced network management in enabling next-generation connectivity.

Utilizing Convolution Neural Network (CNN) Algorithm for the Classification of Visual, Auditory, Read/Write, and Kinesthetic (VARK) Learning Styles Based On Real-Time Datasets

2025-12-26T07:03:29+00:00

Identifying learners’ preferred learning styles is essential for effective personalization in educational environment. The VARK model (Visual, Auditory, Read/Write, and Kinesthetic) is widely used for this course, yet traditional questionnaire-based assessments struggle with scalability, static data, and limited adaptability. This study introduced an optimized Convolutional Neural Network (CNN) framework for real-time, automated VARK classification using multimodal interaction data. Learner engagement was tracked through event listener technique within a learning management system, capturing HTTP+play/pause for visual and auditory media, HTTP+scroll for reading/writing materials, and HTTP+focus/blur for kinesthetic activities. These event listeners were used to track time spent in each modality and combined with corresponding quiz performance scores to form a comprehensive dataset. The CNN model was trained on twelve thousand (12,000) collected datasets of learners from Hunter e-Academy (He-A) learning management system to classify individual learning styles, enabling dynamic adaptation of content delivery.To evaluate performance, the CNN model was compared through A/B testing against other machine learning (ML) models, including Support Vector Machines (SVM), Random Forest, Naive Bayes, and XGBoost. Metrics such as accuracy, precision, recall, and F1-score were used for assessment. The CNN achieved an accuracy of 99.05%, surpassing SVM (98.01%), XGBoost (98.0%), Random Forest (96.69%), Naive Bayes (96.45%), and Decision Tree (95.98%). It demonstrated perfect precision for Auditory and Read/Write, perfect recall for Visual and Auditory, and F1-scores ?0.98 across all categories, addressing the bias and uneven performance observed in unimodal approaches like KNN (89%). The study confirmed the effectiveness of multimodal data fusion for accurate, objective learning style assessment, offering a scalable, AI-driven alternative to surveys and supporting real-time adaptive learning environments.

Implementation of a Smart Farming Automation System

2025-12-23T19:56:28+00:00

This study presents the design and implementation of a smart farming system that automates irrigation on farmland using an Arduino microcontroller, a Wi-Fi module, and various sensors. The system detects the soil moisture level and determines the optimal time to irrigate crops. It also monitors water levels to prevent overwatering, which can damage root systems. The main objectives of this project are: to develop a robust embedded system for real-time data collection from sensors deployed in agricultural fields, to design a user friendly interface for farmers to remotely monitor and control farming processes, and to implement intelligent systems that automate irrigation based on sensor data. Traditional agricultural practices rely heavily on manual labor and often lack real-time monitoring capabilities, resulting in inefficiencies, resource wastage, and suboptimal yields. Furthermore, unpredictable weather and the demand for precise resource management pose significant challenges. Addressing these issues requires a technologically advanced and integrated approach. The methodology adopted follows Rapid Prototyping and Iterative Model and this involves quickly developing an initial prototype, testing its functionality, gathering feedback, and then iteratively improving the design until the final implementation is achieved. The system was developed and tested to ensure functionality aligned with design specifications. The prototype successfully demonstrated autonomous control of irrigation based on soil moisture readings. In conclusion, smart farming—also known as precision agriculture—leverages technologies such as embedded systems, artificial intelligence (AI), and big data analytics. Through the integration of sensors, GPS, and automated machinery, it enables efficient crop and livestock management while promoting sustainability by reducing waste and conserving water.

Design and Construction of a Wireless Automatic Water Monitoring System

2025-12-23T10:03:09+00:00

Water management involves the planning, development, distribution, and control of the optimal use of water resources in an environment—sourced from boreholes, wells, and other means. Ensuring the sustainability of available water resources has become a critical concern globally, as water remains an essential element for human survival. Radio Frequency (RF) refers to the oscillation rate of electromagnetic radiation or radio waves. In this study, a Wireless Automatic Water Monitoring and Pump Control System was proposed, designed, and implemented to wirelessly monitor water levels in a tank using RF technology and to automatically control the pump operation. Sensors were placed at various levels in the tank to detect water levels at any given time. An embedded system, centred around the PIC16F877A microcontroller, was used to process input signals received via RF from the transmitter module. These inputs were processed through an inverter, and the resulting outputs determined whether the pump was activated or deactivated depending on the tank's water level. The system was tested and evaluated. Results showed that it accurately detected water levels and effectively managed the pump, switching it ON when water was low and OFF when the tank was full

Behavioral Analysis Model for Enhancing Attendees Experiences in Events Through K-Clustering Technique

2025-12-23T08:33:58+00:00

Events management landscape plays an important role in delivering an exceptional attendee experience, from planning to implementation and attendee’s engagement which serves as a critical success element for event organizers. Despite the increasing use of technology in event management, there remains a limited understanding of attendees' behavioural engagement in events either during or after for enhanced attendees’ experiences. This study seeks to bridge the gap and examine attendees’ behavioural segmentation using K-clustering technique for the identification of attendees’ engagement during and after events. This research aims to develop a behavioural analysis model using K-clustering techniques to identify attendees' engagement in events for improved attendees’ experiences. The quantitative research method was used for this research. The designed and model implementation was developed using Python Programming language. The results showed that the attendees’ engagement was clustered into four namely the minimally, multidimensional, highly cognitive and quietly, highly affective and socially engaged. Also, there was no string engagement in terms of the observed age or gender. The elbow performance metrics shows that the four behavioural engagement patterns best represent the data without complexities with the within-cluster sum of squares value of 171820.15 as the inflection point. The silhouette score of 0.37 indicates a decent but not perfect clustering and good enough for the earl-stage attendee segmentation and need further tuning for decision making. The study concludes that event attendees participate more in highly affective and multidimensional segments and that K-clustering technique serve as an important method for understanding both low- and high-involvement of attendees in events

Prediction of Loan Defaulters Using Machine Learning

2025-12-26T20:02:58+00:00

Financial institutions face significant challenges in accurately assessing the risk of loan defaults, which can lead to substantial financial losses and impact overall stability. The primary objective of this study is to develop predictive models that accurately identify potential loan defaulters, enabling lenders to make more informed lending decisions. The study addresses the critical need for more reliable and data-driven credit risk assessment tools by employing logistic regression, random forest, and decision tree algorithms. The research design involves a systematic approach to data collection, preprocessing, feature selection, model development, and evaluation. The dataset, sourced from Coursera's Loan Default Prediction Challenge, includes 255,347 instances and 18 features relevant to loan default prediction. The study employed an under sampling technique to address class imbalance and used train-test split to evaluate model performance. Logistic regression, random forest, and decision tree models were trained and assessed for their predictive capabilities. The results indicate that Logistic regression and random forest models demonstrated superior performance, with accuracy rates of approximately 69% and 68%, respectively. The feature importance analysis revealed key factors influencing loan defaults, such as credit score, loan amount, and employment history.

Igbo Text Named Identity Recognition (NER) System using Natural Language Processing Algorithms: A Review

2025-12-23T06:34:43+00:00

This is a review paper, which is concerned with the recent nature of Named Entity Recognition (NER) for the Igbo language. It is a low-resource language spoken in the Southeastern part of Nigeria. Irrespective of the numerous advancements in NER for high-resource languages, Igbo NER so far remains underrepresented. This is for its unique linguistic challenges, which includes morphological richness and dialect variations. In recent times, frank efforts have been put forward by MasakhaNER and WAZOBIA NER projects to develop NER datasets and models for the Igbo language. The existing datasets are limited in size and domain coverage. For this reason there are needs for high-quality, large-scale, manually annotated NER datasets for real-world deployment. This paper reviews the existing literature works on Igbo NER, highlighting the challenges, creating opportunities and looking into the potential applications of NER in developing Igbo digital assistants, intelligent search, and machine translation. This work aims to contribute to the growth and development of low-resource African NLP with the provision of future research in indigenous language NER

Interoperability Solution for Internet of Medical Things in Telemedicine

2025-12-26T09:00:04+00:00

Restricted interoperability among heterogeneous Internet of Medical Things (IoMT) devices and telemedicine platforms brought immense integration challenges. In addressing this, an API-centric solution was built on RESTful services over .NET Core, complemented by a cross-platform mobile app developed using .NET MAUI. The system facilitated standardized data exchange, end-to-end encryption, and real-time cloud syncing. Synthetic patient dataset-based performance evaluation indicated lesser latency, reduced data loss, and improved scalability in comparison to traditional integration models. The framework provides a modular, extensible approach to seamless IoMT integration in resource-limited environments with future applications potentially including AI-augmented decision support and EHR system integration. Pent provides improvements in response time, data loss reduction, and increased scalability over conventional integration approaches. The proposed solution provides a practical, modular, and scalable method for interoperability in IoMT seamless telemedicine enabling more timely and efficient care through healthcare systems, particularly in resource-limited settings. Future possible extensions would be AI-based decision-making support and EHR system integration at a greater scale. The performance was evaluated by using simulated patient data, taking into account throughput, latency, error rate, and efficiency of integration in devices. Results were remarkable.

A Review of Automated Text Summarization Models on Diverse Datasets: An Evaluation Perspective

2025-12-26T07:29:51+00:00

This paper reviews Automatic Text Summarization which is one of the tasks in Natural Language Processing (NLP). It is driven by speedy increase in textual data across domains. The reviews systematically examined the recent advancements in Extractive, Abstractive and hybrid automatic text Summarization Models between 2019 and 2025 using Preferred Reporting Items for Reviews and Meta-Analysis (PRISMA). Selected and relevant related papers were taken from Elsevier, Google scholar, IEEE Xplorer, ACM digital library, and Springer. After removing duplicates (n=96), 174 irrelevant records were removed to meet the inclusion criteria covering models like BERT (Bidirectional Encoder Representations from Transformers), BART (Bidirectional and Auto Regressive Transformers), T5 (Text-To-Text Transformer), TextRank, LSA (Latent Semantic Analaysis), and PEGASUS (Pre-training with Extracted Gap-sentences for Abstractive Summarization Sequence-to-to-Sequence Models) across Diverse datasets including news, scholarly and technical corpora. Extractive approaches depicted strong lexical accuracy and computational efficiency, whereas transformer-based Abstractive models showed superior semantic coherence but needed higher computational costs. This review paper also highlighted persistent gaps including dataset bias, long-document Summarization, hallucination in generative models, and over reliance on traditional metrics such as ROUGE.The results show the need for cross-domain evaluation, hybrid model integration, and adoption of advanced semantic metrics like BERTScore and MoverScore. Future directions should take into priority cross-domain benchmarks, standardized multi-metric evaluation, hybrid approach exploration and testing for long and multilingual documents. In furtherance, Reproducible Reporting of Computational cost such as GPU-hours and failure modes such as hallucinations will support more practical comparisons.

Development of an Android-Based MedicalBot to Diagnose and Suggest Remedies for Tuberculosis

2025-12-26T06:40:43+00:00

Tuberculosis (TB) is one of the top ten causes of death worldwide and the leading cause of death from an infectious disease. TB is an airborne bacterial infection caused by Mycobacterium tuberculosis, which mainly attacks the lungs. People who have Tuberculosis will have to go to the hospital and in many cases, the availability of the medical specialist cannot be guaranteed. In most cases, when the medical specialist is available, the patient will not be able to afford the charges of obtaining the hospital form and test conducted. The research work focused on the three stages of Tuberculosis which are Exposure, Latent and Active stages Tuberculosis. Agile methodology method was used to carry out the methodology and the programming tools used in achieving this are HTML JAVASCRIPT, CSS PHP and SQL as the database. These tools were used to develop a medical bot page where users can interact with the system; learn more page where user can get more information about Tuberculosis and Patient Data form page where user can register after diagnosing. The system is made flexible, versatile and user-friendly. The application has been tested by various students using Android devices operating system and successful result was confirmed.

Integration of NaijaCaptcha System to an Intelligent PDF Reader with Translator

2025-12-23T10:29:39+00:00

Abstract The intelligent PDF reader with integrated translation capabilities is a software solution designed to enhance the accessibility and usability of PDF documents. This project addresses the challenges faced by users who need to interact with multilingual and scanned PDF files, providing a seamless experience for viewing, extracting, translating, and annotating text. The primary objectives of this project are to develop an application that allows users to upload and view PDFs, extract text using Optical Character Recognition (OCR), translate extracted text into various languages, annotate documents and integration of CAPTCHA system called NAIJACAPTCHA. These features are designed to improve efficiency, accessibility and security for users who need to manage and understand content in different languages and formats. Key features of the application include a PDF viewer for navigating documents, OCR for converting scanned images into editable text, translation of text into multiple languages, and annotation tools for highlighting and commenting on PDFs. The application also includes robust user authentication and authorization mechanisms to protect user data and maintain privacy. The system architecture consists of a user-friendly interface built with React, a backend developed using Flask, OCR capabilities provided by Tesseract, and translation services integrated via the Google Translate API. PostgreSQL is used for data storage, ensuring secure and efficient management of user data and application configurations. This study demonstrates the effective integration of various technologies to create a powerful tool that simplifies the management and interaction with complex PDF documents. The intelligent PDF reader with integrated translation capabilities contributes to a more inclusive and efficient digital environment, addressing the needs of users dealing with multilingual and scanned PDF documents.

Particle Swarm Optimization-Random Forest Weather-Based Crop Yield Prediction Model

2025-12-23T08:56:30+00:00

Crop production is a vital source of food for humans, and improving crop yield requires a deep understanding of crop production processes. It has been proven that increasing crop yield reduces poverty, crop failure risk, increases productivity, and optimizes the value of agricultural land. Many factors affect the amount of crop harvested in a specific area and several studies, mainly in the agricultural context, have been conducted to estimate crop yield production with Machine learning (ML) techniques. This study explores five cereal crop yields: rice, maize, wheat, sorghum, and soybeans with Particle Swarm Optimization (PSO) and Random Forest prediction approaches. Performance metrics such as R2 score, Mean Absolute Error, and Root Mean Squared Error confirm the authenticity of the model. The result of the optimized Crop yield prediction has an R2 score of 97.13, MAE of 124.75, and RMSE of 1273.73. The model performed better than other existing approaches, such as Random Forest (RF) and Decision Tree (DT). This study will provide farmers with reliable crop yield predictions, enabling better planning based on weather conditions.

Enhanced Malaria Detection Model using Deep Convolutional Neural Network with Comprehensive Data Augmentation and Grad-CAM Explainability for Clinical Trustworthiness

2025-12-23T06:53:44+00:00

Malaria remains a major global health challenge, particularly in sub-Saharan Africa and parts of Asia, where accurate and timely diagnosis is essential for effective treatment and control. Traditional microscopic examination, while widely used, is labor-intensive, subjective, and prone to misdiagnosis. To address these limitations, this study proposes deep learning-based approaches for automated malaria parasite detection from thin blood smear images. An enhanced malaria detection model using deep convolutional neural network with comprehensive data augmentation and Grad-CAM was developed. Using the NIH Malaria Dataset comprising 27,514 validated images, the models were trained and tested with rigorous preprocessing, augmentation, and stratified sampling. Results show that the CNN model achieved 96.37% accuracy, 98.40% recall for parasitized cells, and an AUC of 0.9935, outperforming conventional methods and providing robust generalization for unseen data. This study highlights the potential of deep learning in advancing malaria diagnostics while also addressing critical deployment considerations, including error calibration and clinical applicability. This enhances clinical Trustworthiness.

Design and Implementation of a Blockchain-Based Certificate Verification System for Secure Academic Credential Authentication

2025-12-26T09:11:50+00:00

Certificate forgery is a pervasive issue in Nigeria’s educational system, undermining trust in academic credentials and causing delays in verification processes. Traditional paper-based systems are inefficient, costly, and susceptible to tampering. This study presents a blockchain-based certificate verification system that leverages Ethereum smart contracts, InterPlanetary File System (IPFS) for decentralized storage, and PostgreSQL for off-chain metadata management to provide a secure, tamper-proof, and real-time verification platform. The system, implemented using React.js for the frontend, Node.js for the backend, and Solidity for smart contracts, enables institutions to issue digital certificates with embedded QR codes and allows instant verification by employers and other stakeholders. Testing on the Ethereum testnet demonstrated 98% accuracy in detecting forged certificates, with verification times under 2 seconds. The system enhances transparency, reduces administrative overhead, and aligns with Nigeria’s push for technological innovation in education. Challenges such as Ethereum gas fees and institutional adoption are discussed, with recommendations for scalability and mobile support.