University of Ibadan Journal of Science and Logics in ICT Research

A Machine Learning-Based Fraud Prevention Model for Improving Customers’ Trust in E-Commerce

journal manager — Fri, 07 Mar 2025 00:00:00 +0000

The growth of e-commerce has led to significant challenges regarding fraud, resulting in a decline in customer trust and confidence in online transactions. This research proposes a comprehensive Fraud Prevention Model aimed at enhancing customer trust and security within e-commerce platforms by integrating advanced machine learning (ML) techniques, an Address Verification System (AVS), and Two-Factor Authentication (2FA). The model leverages Convolutional Neural Network - Long Short-Term Memory Network (CNN-LSTM) and Random Forest techniques to capture the complexities and temporal dependencies of e-commerce transaction data. The AVS component of the system verifies transaction legitimacy by comparing billing addresses with credit card records, and the implementation of 2FA adds an extra layer of security. The system's effectiveness was evaluated through rigorous testing using a dataset of transaction records. The results indicate that the combined approach of machine learning, AVS, and 2FA significantly enhances the detection of fraudulent transactions and improves overall customer trust in e-commerce platforms.

A Systematic Review of Computational Approach to Pipeline Leakage Detection in a Water Distribution Network

O.S. Ojo, R.O. Akinyede, O.D. Alowolodu, A Adebayo — Fri, 07 Mar 2025 00:00:00 +0000

The detection and localization of leakages in water distribution networks is crucial for both the conservation of
resources and the efficient operation. The process network has proved to be a difficult task over years,
considering the complexities inherent in water distribution networks. The enormous interconnected pipelines
make the leakage detection and location process burdensome. Computational techniques play a significant role in
this domain by offering advanced tools and techniques for leakage detection. This study, therefore, performed a
systematic review of published articles on computational leakage detection and localization in a water
distribution network. Findings show the number of recent quality studies on the computational approach to water
distribution network leakage research is beginning to dwindle, considering the journal's impact factor. In the
recent studies, a deep learning algorithm is beginning to trend as the most significant computational technique,
as it accounts for 13.21 % (n=7) of the pipeline leakage research output. The univariable predicated studies
account for 83.33%of the research output disseminated in the past five years. The invention of various efficient
learning methods and network structures in deep learning algorithms makes it suitable for the realization of
multi-disciplinary studies, as the multi-variable concept will reduce false positives and negatives, enhancing the
overall reliability of leak detection and localization models in future studies.

Framework for a Stimulated Predictive Distributed Learning Method

E.C Igodan, J.O Iyekowa — Fri, 07 Mar 2025 00:00:00 +0000

Due to the intrinsic properties of high-dimensional microarray datasets, most feature selection approaches do not

scale well, which makes these models inapplicable and impairs the performance of most classifiers. This study

used data complexity and stability measures to maintain class distribution and reduce features variability while

proposing a novel predictive distributed FS model through horizontal partitioning. Brain tumour microarray

benchmark was employed for implementation. Six classifiers as well as feature selection methods were

employed along with their ensemble learning techniques. The study observed the proposed distributed model

with an average accuracy of 98.54% and 99.67% obtained from both the single and ensemble

models respectively.

Improved Stock Price Prediction Model in the Nigeria Bank Sector Using Ensemble Machine Learning Models

Ayoade Akeem Owoade — Fri, 07 Mar 2025 00:00:00 +0000

Stock market prediction remains a critical challenge in emerging economies, particularly within volatile

financial landscapes like Nigeria. Despite significant technological advancements, existing research

predominantly relies on single-model approaches that inadequately capture the complex, non-linear dynamics of

financial markets. This study addresses the methodological gap by developing an ensemble machine learning

model for predicting stock prices in the Nigerian banking sector. The research utilized historical stock price data

from Guaranty Trust Bank and First Bank (2018-2023), integrating advanced preprocessing techniques,

employing rigorous data transformation, feature standardization, and cross-validation strategies, the study

transforms raw financial data into a robust predictive framework. Empirical results reveal distinct performance

metrics across ensemble models: Among the models, Gradient Boosting achieved an MAE of 0.1547, MSE of

0.0918, and RMSE of 0.999, while the Stacking Regressor yielded an MAE of 0.1912, MSE of 0.1396, and

RMSE of 0.9989, highlighting their accuracy and reliability in volatile market conditions. The ensemble

methodology demonstrates superior performance in capturing intricate market dynamics, offering significant

improvements over traditional forecasting techniques by integrating macroeconomic indicators and advanced

machine learning algorithms. The findings underscore the potential of ensemble machine learning in decoding

complex financial patterns, providing valuable insights for investors, financial analysts, and policymakers.

Integrating Zero-Trust Architecture with Deep Learning Algorithm to Prevent Structured Query Language Injection Attack in Cloud Database

M. E Timadi , E.C.M Obasi — Fri, 07 Mar 2025 00:00:00 +0000

The increasing reliance on cloud databases has made them a prime target for cyber attacks, with Structured

Query Language (SQL) injection being a particularly devastating threat. SQL injection attacks pose significant

threats to database security, compromising sensitive information. Deep learning algorithms have emerged as

effective solutions to detect and prevent SQL injection attacks. This study proposes a novel approach to

detecting SQL injection attack by integrating deep learning-based detection with zero-trust architectute. The

proposed system utilizes a Feed-Forward Neural Network (FNN)to analyze database queries and detect potential

SQL injection attacks. The FNN model is trained on a dataset of labelled queries, allowing it to learn patterns

and anomalies indictive of SQL injection attacks. The output of the FNN model is then integrated with zero-

trust architecture, which enforces strict access controls and authentication mechanisms based on the results

generated by the FNN model. The model exhibits a precision score approximating 100% accuracy in the

classification of queries deemed normal, while achieving a 94% rate of correct classification for queries

indicative of SQL injection attacks. By leveraging advanced machine learning techniques, our approach aims to

identify and block malicious queries in real-time, ensuring the integrity and security of cloud-based data.

Through a comprehensive evaluation, we demonstrate the effectiveness of our deep learning-based solution with

zero-trust architecture in detecting SQL injection attacks with high accuracy and low false positives.

Predicting Students’ Academic Performance in Virtual Learning Environment Using Pearson Correlation Coefficient

F.O Adelodun , W Sakpere, K. F Famurewa , Y.J Oguns — Fri, 07 Mar 2025 00:00:00 +0000

Feature Selection involves selecting the most relevant features from a dataset during the prediction process. The

selection method of features greatly influences how accurate, understandable, and effective predictive models

are. Predicting students' academic success or struggle in a Virtual Learning Environment (VLE) is limited.

Students who drop out of online courses are substantially more numerous than those who drop out of traditional

courses [1,2]. The methodology followed in the study involved the use of two approaches: training and testing

machine learning models with features selected from the dataset, and the second approach involved training and

testing the machine models using all features in the dataset without feature selection. The Pearson Correlation

Coefficient (PCC) feature selection method is used to select the features used for prediction. The two

approaches were compared in terms of their impacts on the performance of the machine learning algorithms.

The study was carried using nine classification models, which include Logistic Regression, K-Nearest

Neighbour (KNN), Support Vector Machine (SVM), Random Forest, Gradient Boosting, XGBoosting,

LightGBM, MLP classifier (Neural Network) and Naïve Bayes. The result of the study showed that logistic

Regression show highest accuracy mean of 0.7333 with feature selection and reduced accuracy mean of 0.7188

when all features were used in the prediction process. Without feature selection, the accuracy mean of Random

Forest is 0.6813 and applying PCC feature selection to select the features for prediction, the accuracy mean of

Random Forrest increased to 0.7333 revealing that feature selection method such as PCC is important for

improving model performance.

A Machine Learning-Based Predictive Model for the Classification of Academic Performance of Students

Fri, 07 Mar 2025 00:00:00 +0000

Predicting student academic performance is critical for enhancing personalized learning and improving educational

outcomes. Traditional assessment methods, while useful, often fail to capture the complex factors influencing

performance, such as socio-economic background and engagement metrics. This study explores the development of a

predictive model using an ensemble of machine learning algorithms to classify students' academic performance in

higher institutions. By leveraging data collected from Department of Computer Science, Tai Solarin University of

Education records, relevant features were selected using the mutual information method. The ensemble model was

formulated and simulated using multiple machine learning algorithms such as Naïve Bayes (NB), Support Vector

Machines (SVM) and Decision Trees (DT) in the Google CoLaboratory environment. The model’s predictive accuracy

was evaluated based on key performance metrics, including accuracy, precision, and F-measure. Results indicate that

the ensemble approach outperforms single-model methods by enhancing prediction robustness and reducing variance.

This study demonstrates the effectiveness of machine learning techniques in identifying at-risk students early with NB

and SVM having 100% accuracy respectively, allowing for timely interventions and improved resource allocation.

Moreover, it contributes to evidence-based decision-making in educational institutions, helping to optimize learning

experiences and boost student retention rates.

A Comparative Analysis of Ensemble Machine Learning Algorithms for Bank Customer Churn Prediction

journal manager; Gbenga. O Ogunsanwo — Fri, 07 Mar 2025 00:00:00 +0000

Customers churn became a serious issues to banks manager because customers have numerous options where to

save their money. This justify why many researchers are attracted to this area. This study developed a bank

customers churn predictive model. The study used dataset from kaggle.com repository. It consists of 10127

instances and 20 parameters. One Hot Encoder was used as data preprocessing on the dataset. The data was divided

into 80% for training and 20% for testing. The predictive model was created using Long Short-Term Memory

(LSTM), Ensemble LSTM, and Random Forest (RF). The results of the model revealed LSTM with F1 score of

0.94, accuracy of 0.9235, specificity of 0. 6635 sensitivity of 0.97, AUC of 0.95 and loss value of 0.1663. Ensemble

LSTM with F1 score of 0.94, accuracy of 0.9057, specificity of 0.554, sensitivity of 0.98, AUC of 0.92 and loss

value of 0.238. RF with F1 score of 0.97, accuracy of 0.95, specificity of 0. 774, sensitivity of 0.99, AUC of 0.99

and loss value of 0.15. The study concluded that RF outperformed both LSTM and Ensemble LSTM. Also pointed

out that customer’s gender, marital status, customer income category and age against attrition are determining factor

for customer churn prediction. The model is recommended for banking sector to assist in decision making. Future

work can be done using more ensembles techniques and perform more data expository

A Multiclass Model for Adversary Domain Name Classification using Tree Based AI Classifiers

B. B Odigie, O. P Bernard — Fri, 07 Mar 2025 00:00:00 +0000

The rising prevalence of AI-generated adversary (malicious) domain names has escalated the challenge of
combating cybercrime, particularly as spamming, phishing, and malware activities become increasingly common
online. Traditional approaches, such as blacklisting, binary detection systems, and basic lexical analysis of domain
names, prove insufficient for real-time identification of malicious domains across various cyber threat landscapes.
This study presents a comprehensive strategy for the multiclass detection of malicious domain names (MDNs)
utilizing data mining techniques. It investigates feature engineering processes, including dimensionality reduction
and variance inflation factor analysis, to identify and select domain name features that enhance the performance of
advanced AI and machine learning classifiers in classifying MDNs. We employed a train/test split ratio and cross-validation methods on the CIC-Bell-DNS2021 public dataset for training some cutting-edge AL/ML classifiers. The
findings reveal that tree-based machine learning algorithms, particularly the Extreme Gradient Boosting (XGBoost)
algorithm achieved outstanding results, with a mean accuracy score of 0.9998 (100%). Additionally, regarding
execution time, XGBoost displayed a notable advantage, requiring less time to build models, which could
significantly influence real-time detection capabilities when implemented as a cybersecurity tool for detecting
malicious domain names.

Optimising Malaria Prediction from Cell Images Using Forward Selection and Support Vector Machine Classifier

O Folorunsho, O. O Faboya, S. A Mogaji, E Willie, I Ochidi — Fri, 07 Mar 2025 00:00:00 +0000

Malaria is a significant health concern, primarily affecting tropical and subtropical regions. Traditional
diagnostic methods for malaria detection, such as microscopic blood smear analysis of cell images, are time
consuming, dependent on trained specialists, and prone to variability. Timely and accurate malaria detection is
crucial for prompt treatment and preventing severe complications. Therefore, this study developed a machine
learning (ML)-based model that could accurately predict malaria by analysing microscopic cell images,
enabling efficient and reliable diagnosis to support timely treatment decisions. Using the Kaggle malaria
dataset comprising 26,159 blood smear images, this study uniquely integrates forward feature selection and
Support Vector Machines (SVM) to enhance malaria prediction accuracy. Unlike existing works, it addresses
gaps in transparency and reproducibility in feature selection methods used for high-dimensional medical image
datasets. Forward selection was employed to optimise and select relevant features for the model, reducing
computational complexity and enhancing its performance. The SVM model achieved an accuracy of 97.1%,
recall of 97.4%, precision of 96.8%, F1-score of 96.9%, and an AUC score of 97.4%. These findings highlight
the potential of ML in automating malaria detection and demonstrate the practical advantage of combining
feature selection with high-performing classifier to optimise diagnostic workflows, especially in resource
limited settings.

Evaluation of Machine Learning-Based Algorithm to Predicting Loan Default in Nigeria

K. O Efekodo, O. S Akinola, A. A Waheed — Fri, 07 Mar 2025 00:00:00 +0000

Accurately predicting loan defaults is critical in the financial sector to minimize losses and optimize credit risk
management. Traditional creditworthiness assessment methods often fail to capture the complex, dynamic
interactions in financial data, leading to inaccurate predictions. This study harnesses advanced machine learning
techniques to enhance the prediction of loan defaults, aiming to outperform traditional statistical models. A
dataset containing 50,000 borrower records with diverse characteristics, including demographic, financial, and
loan-specific features, was utilized. The data was split into training (70%) and test (30%) sets for model
development and evaluation. Various machine learning algorithms were tested, including Logistic Regression,
Decision Trees, Gradient Boosting Classifiers, Random Forest, and Gaussian Naive Bayes. The Gaussian Naive
Bayes (GaussianNB) model demonstrated superior performance, achieving an accuracy of 78.8% on the test set.
This model effectively captured complex patterns in the high-dimensional data, significantly reducing false
positives and false negatives compared to other models. The findings suggest that machine learning models,
particularly GaussianNB, offer substantial improvements in predictive accuracy for loan default risk assessments.
This findings can enhance lenders' decision-making processes by improving risk stratification and resource
allocation. Future research should explore integrating non-traditional data sources, such as behavioral and
macroeconomic variables, and employing deep learning techniques to further refine predictive accuracy.

Leveraging Artificial Intelligence for Detecting Insider Threats in Corporate Networks

J. T Nnodi, E. C. M Obasi — Fri, 07 Mar 2025 00:00:00 +0000

In the modern corporate environment, insider threats pose a significant risk to data integrity, financial stability,
and overall cybersecurity. Unlike external attacks, insider threats originate from individuals within an
organization like employees, contractors, or partners who possess legitimate access to critical systems.
Traditional security measures often fail to identify these threats due to the complexity of distinguishing malicious
behaviour from regular activities. Artificial Intelligence (AI) based systems, with their ability to analyse large
datasets, detect subtle patterns, and adapt to evolving threat landscapes, offer a powerful approach to insider
threat detection. This research involves the application of machine learning algorithms to identify deviations from
normal users’ activities in corporate networks. The methodology involves analysing user behaviours and access
patterns, development and training a machine learning model for classifying user behaviours into normal or
abnormal activity. The system helps to identify abnormal user activities and flags suspicious activities in real time,
providing an early warning sign for potential breaches. The results demonstrate the effectiveness of machine
learning in enhancing threat detection, reducing insider threats, and improving overall cybersecurity in corporate
networks.

A Review of Fixed Input Size Limitation in Convolutional Neural Networks Models and Proposed Solutions

K.F Famurewa, W Sakpere, F.O Adelodun — Fri, 07 Mar 2025 00:00:00 +0000

Convolutional Neural Networks (CNNs) are incredibly powerful deep learning techniques that have been applied
to computer vision applications to yield innovative results. CNNs are ideal for applications like object
identification, image segmentation, and image classification because they can automatically extract pertinent
information from the images without human supervision. While CNNs can attain state-of-the-art performance in
many applications and domains, most CNNs currently have limitations in training and prediction due to their
sensitivity to image size. As a result, image recognition datasets are typically downsized to the input size
specification of the CNN models. This study's objective is to examine CNN models and suggest possible
solutions to tackle the fixed input size problem that exists in CNN models.

Education Systems Interoperability: Implications for Privacy and Security in Educational Management Information Systems

O, A Adenubi, N Samuel, A. O Oyenuga — Fri, 07 Mar 2025 00:00:00 +0000

The importance of robust Educational Management Information System (EMIS) becomes very essential in
addressing the complexities of data management, particularly as Nigerian educational systems is progressively
leveraging the interrelated platforms in order to enhance operational efficiency and data sharing. However,
educational sector is faced with several related challenges like: fragmented data, management systems, data
privacy concerns, inadequate technology infrastructure, interoperability that is poor between different platforms,
and lack of standardized protocols. These challenges made most institutions of learning to compromise the
security and integrity of sensitive information. This study presents conceptual model that exemplifies the
synergistic interactions among the major components of EMIS, by stressing the main roles of interoperability,
security and privacy. Interoperability is central and surrounded by Data Governance, Technology Infrastructure,
Privacy Measures, Regulatory Compliance, Security Measures, and Stakeholder Engagement. Each of the
components is interconnected to illustrate how technology infrastructure enable effective data exchange while
preserving sensitive information. The integration of these components in the proposed model offers qualitative
understanding for educational institutions to strive well in enhancing EMIS while securing stakeholder’s privacy.
This strategy addresses the present limitations of EMIS in Nigeria, opening way for a more efficient and safe
educational data management system. It is recommended that educational institutions should be encouraged in
adopting standardized procedures for seamless integration of EMIS in order to facilitate overall functionality
and efficient exchange of data. Also, institutions should develop clear data governance policies that prioritize
privacy and security, ensuring regulatory compliance and promoting responsible data use through regular
training and awareness programs.

A Comparative Evaluation of Embedding Techniques from Language Models for Automatic Grading of Short Answer Questions

V Odumuyiwa, O Adewoyin, A Fagoroye, E Fasina, B Sawyerr, O Sennaike — Fri, 07 Mar 2025 00:00:00 +0000

An automatic grading system of short answer questions on an e-learning platform can help reduce stress, save
time, increase the productivity of instructors and help provide feedback to students in record time. However, the
success of automatic grading of short answer questions (open-ended questions) depends on the ability of the
computer to adequately capture the semantic similarity between students’ answers and the reference answer
provided by the examiner. This paper presents a comparative study of some embedding techniques from
language models for automatic grading of short answer questions in order to address the longstanding challenge
of automating the assessment of students' responses to open-ended questions. It studies five embedding
techniques such as Word2vec, Bi-LSTM, BERT, SBERT, and OpenAI on four datasets (SemEval, Texas,
ASAG, and MIT) to find the best method among them for Automatic Short Answer Grading (ASAG).
Experiments include regression tasks and classification tasks using Mean Squared Error (MSE), Pearson
Correlation, and accuracy as metrics for evaluation. The results indicate that fine-tuned BERT achieved the
highest accuracy of 75% on SemEval dataset in classification tasks, while OpenAI performed better in the
regression tasks with a MSE of 0.57 on the Texas dataset. The research highlights automated grading as a means
to reduce instructors' workload while enhancing the quality of feedback provided to learners. Future studies will
focus on extending the experiments to include both domain-specific and non-domain-specific.

A Framework for Education Technology Integration in Nigerian Basic School System: Digital Framework for Technology Integration in Education (DiFTIE) for Basic School System

O. A Adenubi, N Samuel, A. O Oyenuga — Fri, 07 Mar 2025 00:00:00 +0000

The integration of technology into education has gained significant attention globally, however, the existing
frameworks such as: Technological Pedagogical Content Knowledge (TPACK), Substitution, Augmentation,
Modification, Redefinition (SAMR) and (Technology Integration Matrix (TIM) failed to address the unique
challenges faced by technology resource-limited contexts like Nigeria. This study designed and developed a
contextual model termed Digital Framework for Technology Integration in Education (DiFTIE) to address the
unique challenges facing technology integration in Nigerian basic schools. It was also developed to bridge the
existing digital divide among students by improving educational performance of students from diverse socio
economic backgrounds and to promote equitable access to technology-enhanced learning. The DiFTIE
Framework fills a major gap by suggesting a realistic, actionable, and adaptable model that is tailored to
Nigeria's socio-economic realities. DiFTIE framework surpasses traditional frameworks such as TPACK,
SAMR and TIM by emphasizing on policy alignment, foundational readiness and community involvement
(major elements in resource-limited contexts to enhance sustainable integration of technology). Components of
DiFTIE Framework include developing localized educational content, enhancing ICT infrastructure, provision
of teacher training programs and strengthening policy support. The DiFTIE Framework also provides well
structured and sustainable strategies for integrating educational technology into pedagogic experiences due to
the fact that the framework recognizes the specific needs of Nigerian educational system and its challenges.
Therefore, it is recommended that implementing the DiFTIE Framework would promote equitable access to
technology-enhanced learning for all students irrespective of gender and socio-economy background, display a
transformative role in bridging the digital divide in Nigerian basic education and to improve educational
performances of students across diverse socio-economic backgrounds.

Enhancing Symmetric Encryption Using Digital Signatures

Y. Y Chamo — Fri, 07 Mar 2025 00:00:00 +0000

Maintaining the confidentiality and integrity of digital documents transmitted through electronic media is a
critical security concern in the field of Information Security. To address this security concern, this paper
proposes a system that uses a digital signature to ensure the authenticity, non-repudiation and integrity of the
transmitted data and it also uses symmetric encryption to provide authentication and confidentiality of the
transmitted data. The Rivest, Shamir & Adleman (RSA) algorithm was used to implement the Digital Signature
while the Advanced Encryption Standard (AES) was used for symmetric encryption. The system involves
encrypting a plaintext using AES, then a hash function (SHA-256) is used to create a hash value of the
ciphertext and the private key of the RSA algorithm is used to encrypt the hash value to produce the digital
signature. The ciphertext and the digital signature are attached and sent to the recipient. The digital signature is
decrypted by the recipient to obtain the hash value of the ciphertext, then it verifies if it is a valid signature
before proceeding to decrypt the ciphertext using the AES secret key. The proposed system was evaluated
against the existing AES algorithm. The size of the test file was observed and analyzed before and after
encryption, this showed that the size did not change. Different RSA key sizes were used to perform signature
and verification processes to see how long it takes to perform the operations, this also showed that the smaller
the key size the faster the signature and verification processes and the verification process is a much faster
process than the signature process. The system was able to meet the cryptography objectives and will be useful
to individuals and businesses in transmitting sensitive information over insecure communication mediums.

Development of a Knowledge Management System to Support Intelligent Rice Farming

Fri, 07 Mar 2025 00:00:00 +0000

Rice is a staple food worldwide. It is most common in Asia, Africa and Latin America. In Ghana, the annual rice
consumption is about 1.5 million metric tons. However, about 60% of this demand is imported majorly from
Asia. This high reliance on imported rice is one of the contributory factors that weaken Ghana’s foreign
exchange reserves. Hence, there is a need to encourage local rice production. However, the efforts by local
farmers are thwarted by many challenges, including pest infestations, bird interference, insufficient technology
for efficient fertiliser and herbicide applications, and the absence of reliable systems for predicting rainfall
patterns. Additionally, inadequate access to modern agricultural extension services and the lack of advanced
storage facilities exacerbate these difficulties. Although numerous intelligent agriculture systems exist that could
address these issues in an environmentally sustainable manner, farmers in this region remain largely unaware of
such technologies and persist with outdated and inefficient methods. This study sought to address these
challenges by developing a customised Knowledge Management System (KMS) aimed at facilitating knowledge
dissemination and supporting intelligent agricultural practices in rice farming. The research used a system
prototyping methodology to produce a prototype KMS, which the System Usability Scale (SUS) evaluated for
usability. The system achieved an average score of 70.025, surpassing the threshold for "Acceptability
Usability," which denotes that the KMS meets the minimum standards for practical application. This result
highlights the potential for the KMS to enhance agricultural practices and improve productivity within the
community.

Assessment of the Initial Adoption and Implementation of an Electronic Medical Records (EMR) System in a Nigerian Teaching Hospital

K. F Famurewa, W Sakpere, F. O Adelodun — Fri, 07 Mar 2025 00:00:00 +0000

All medical information related to patients can be displayed in a digital system known as the Electronic Medical
Record System. The system is an attempt by many medical facilities to design and implement a means of collecting,
storing, and presenting patients’ information at the point of care. EMRs are implemented worldwide because they
are recognized as a potential means of improving medical services' quality, safety, and efficiency. In Nigeria, the
Federal Ministry of Health has acknowledged the significance of EMRs for quality enhancement in healthcare
delivery and has directed all teaching hospitals to computerize their clinical processes. Based on this directive, some
teaching and national hospitals are now computerized. This paper aims to assess the initial adoption and
implementation of the EMR system in a Nigerian teaching hospital. The research design adopted in this paper is a
qualitative research design. Primary data were collected via participant observation and face-to-face interviews with
medical staff and information technology personnel who deployed the system. Also, the researchers reviewed
existing literature on adoption of electronic medical records systems and the pros and cons of adopting an electronic
medical records system. The tools used in developing the EMR system were JavaScript, C#, MSSQL server, Web
server (IIS), and Web browser is the program the user uses to view the web pages. The system has many modules
which can only be accessed by authorized members of staff based on the role they perform. The system comprises
of patient information module, Billing and payment module, Nurses module, Doctors module/clerking module,
Customizable Reporting/Quality reporting, clinical coding, NHIS module, Diagnoses, Diagnostic imaging system
(Radiology), E-prescription, and Laboratory information system. The system can still be improved upon through the
inclusion of more relevant modules for better clinical and patient experience.

Artificial Intelligence-Assisted Retrieval of Owo Cultural Artifacts from Heterogeneous Online Sources

B. R Ijasan, O Oriola, E. O Oyekanmi — Fri, 07 Mar 2025 00:00:00 +0000

Preservation of cultural heritage has increasingly been dependent on artificial intelligence (AI) to tackle issues of
documentation, accessibility and retrieval. The present research centers around the system of an AI-based
retrieval system designed for Owo cultural objects, a historically important collection in Ondo State, Nigeria.
Through machine learning, natural language processing, and chatbot technology, the system overcomes access
barriers, improves familiarity with, and comprehension of these artifacts. The study is based on data collection
across a range of sources, data pre-processing steps to enable structured storage, and the development of a Flask
based API to provide a platform for easy and on demand retrieval. A chatbot driven by Botpress is used as the
user interface to allow the system to be used via natural language queries. The AI model, by learning textual and
image-based representation, showed excellent accuracy for artifact retrieval, and multimodal learning itself
further improved performance of classification. The paper demonstrates the possibility of application of AI as
connecting bridge between classical preservation techniques and current digital accessibility, guaranteeing the
permanent recording and interaction with Owo cultural properties. Future developments include augmenting
NLP functionality, scaling the system, and increasing the scope of the datasets in order to further improve
accuracy of artifact retrieval.

A Robust Biometric Authentication Framework for Access Control

T Achimba, O. J Okhuoya, R. O Akinyede, P. A Alabi, A Ibrahim, G Ateata — Fri, 07 Mar 2025 00:00:00 +0000

Unauthorized access poses significant security concerns, jeopardizing the confidentiality, integrity, and
availability of critical data and resources. Ensuring authorized access is essential for protecting sensitive systems
across diverse fields, including smart buildings, military bases, hospitals, airports, and financial institutions.
Biometric authentication has emerged as a reliable solution for access control, leveraging unique human traits for
verification. However, traditional feature-based biometric systems are limited by environmental sensitivity, poor
generalization, and vulnerability to spoofing, while deep learning-based systems face challenges such as high
computational demands, reliance on large datasets, and lack of interpretability. To address these limitations, this
research proposes a hybrid biometric authentication framework that combines the strengths of deep learning,
specifically Residual Network (ResNet)-a Convolutional Neural Network (CNN), with the Local Binary Pattern
(LBP) method. By integrating interpretable, computationally efficient features from LBP with ResNet’s ability to
learn complex patterns, the framework improves robustness, reduces overfitting, and enhances scalability. This
approach offers a balanced, efficient solution for secure biometric authentication, tailored for real-world and
resource-constrained environments.