DSpace Angular :: Browsing by Author "Isiaka, Mope Rafiu"

Browsing by Author "Isiaka, Mope Rafiu"

Now showing 1 - 20 of 24

A chi-square-SVM based pedagogical rule extraction method for microarray data analysis
(International Journal of Advances in Applied Sciences (IJAAS), 2020-06-10) Mukhtar Damola Salawu; Micheal Olaolu Arowolo; Sulaiman Olaniyi Abdulsalam; Isiaka, Mope Rafiu; Bilkisu Jimada-Ojuolape; Mudashiru Lateef Olumide; Kazeem A. Gbolagade
Support Vector Machine (SVM) is currently an efficient classification technique due to its ability to capture nonlinearities in diagnostic systems, but it does not reveal the knowledge learnt during training. It is important to understand of how a decision is reached in the machine learning technology, such as bioinformatics. On the other hand, a decision tree has good comprehensibility; the process of converting such incomprehensible models into an understandable model is often regarded as rule extraction. In this paper we proposed an approach for extracting rules from SVM for microarray dataset by combining the merits of both the SVM and decision tree. The proposed approach consists of three steps; the SVM-CHI-SQUARE is employed to reduce the feature set. Dataset with reduced features is used to obtain SVM model and synthetic data is generated. Classification and Regression Tree (CART) is used to generate Rules as the Last phase. We use breast masses dataset from UCI repository where comprehensibility is a key requirement. From the result of the experiment as the reduced feature dataset is used, the proposed approach extracts smaller length rules, thereby improving the comprehensibility of the system. We obtained accuracy of 93.53%, sensitivity of 89.58%, specificity of 96.70%, and training time of 3.195 seconds. A comparative analysis is carried out done with other algorithms.
A Framework for Dropout Monitoring and Management System
(Computing, Information Systems, Development Informatics & Allied Research Journal, 2017-07-08) Isiaka, Mope Rafiu
Dropping out of school has economic and social implications on the dropouts and the society at large. Being a global problem, efforts of researchers on the causes, effects, preventives and control methods are well documented in the literature. However, despite these efforts, the world is still witnessing unacceptable rate of school dropouts. The National Dropout rate in United State is 8.1% , that of the foreign borns is 20.7%. Though, there is no coordinated dropouts figure in Nigeria, that of South Africa is 40%. Presently, in Africa, issue of dropouts is not receiving enough attention of the people and government. It is being viewed as the problem of the victims and their families. Hence, available researches on the subject matter are international. The techniques and findings reported cannot therefore be generalized or directly adopted in Nigeria for societal, cultural, social, and political diversities. This work presents a conceptual framework for dropout monitoring and control in Nigerian Universities. Theories of Attribution, Social- Capital, and Engagement provide the bases for the design. The causes, prevention and control components of the framework are explained. The significance of the Early Dropout Warning and Control subsystem is stressed and techniques for its realization are also explained. On implementation, the dropout risk factors in Nigeria shall be determined and the most appropriate Data Minning and Artificial intelligence techniques for the subsystem shall be discovered. Also, the acceptability and effectiveness of the system shall be evaluated. Inferences drawn and the lessons learnt shall form the basis for further recommendations and future works.
A Hybrid Dimensionality Reduction Model for Classification of Microarray Dataset
(I.J. Information Technology and Computer Science, 2017-11-08) Abdulsalam, Olaniyi Sulaiman; Isiaka, Mope Rafiu; Gbolagade, Alade Kazeem
In this paper, a combination of dimensionality reduction technique, to address the problems of highly correlated data and selection of significant variables out of set of features, by assessing important and significant dimensionality reduction techniques contributing to efficient classification of genes is proposed. One-Way ANOVA is employed for feature selection to obtain an optimal number of genes, Principal Component Analysis (PCA) as well as Partial Least Squares (PLS) are employed as feature extraction methods separately, to reduce the selected features from microarray dataset. An experimental result on colon cancer dataset uses Support Vector Machine (SVM) as a classification method. Combining feature selection and feature extraction into a generalized model, a robust and efficient dimensional space is obtained. In this approach, redundant and irrelevant features are removed at each step; classification presents an efficient performance of accuracy of about 98% over the state of art.
A Hybridized Feature Extraction Model for Offline Yorùbá Document Recognition
(Asian Journal of Research in Computer Science, 2023-05-16) Ajao, Jumoke Falilat; Isiaka, Mope Rafiu; Babatunde, Ronke Seyi
Document recognition is required to convert handwritten and text documents into digital equivalents, making them more easily accessible and convenient to store. This study combined feature extraction techniques for recognizing Yorùbá documents in an effort to preserve the cultural values and heritages of the Yorùbá people. Ten Yorùbá documents were acquired from Kwara State University’s Library, and ten indigenous literate writers wrote the handwritten version of the documents. These were digitized using HP Scanjet300 and pre-processed. The pre-processed image served as input to the Local Binary Pattern, Speeded-Up-Robust-Features and Histogram of Gradient. The combined extracted feature vectors were input into the Genetic Algorithm. The reduced feature vector was fed into Support Vector Machine. A 10-folds cross-validation was used to train the model: LBP-GA, SURF-GA, HOG-GA, LBP-SURF-GA, HOG-SURF-GA, LBP-HOG-GA and LBP-HOG-SURF-GA. LBP-HOG-SURF-GA for Yorùbá printed text gave 90.0% precision, 90.3% accuracy and 15.5% FPR. LBP-HOG-SURF-GA for Handwritten Yorùbá document showed 80.9% precision, 82.6% accuracy and 20.4% (FPR) LBP-HOG-SURF-GA for CEDAR gave 98.0% precision, 98.4% accuracy and 2.6% FPR. LBP-HOG-SURF-GA for MNIST gave 99% precision, 99.5% accuracy, 99.0% and 1.1% FPR. The results of the hybridized feature extractions (LBP HOG-SURF) demonstrated that the proposed work improves significantly on the various classification metrics.
A Hybridized Honey Encryption for Data Security using Residue Number System
(International Journal of Information Security and Cybercrime, 2023-09-12) Giwa, Afoluwaso Tawakalitu; Abiodun, Oludare Isaac; Omolara, Abiodun Esther; Isiaka, Mope Rafiu
Different encryption algorithms such as Advanced Encryption Standard (AES), Rivest, Shamir, Adleman (RSA) were proposed to protect the privacy of the data. However, most of these existing methods are vulnerable to a brute-force attack because the cipher text remains unintelligible until the original data is found. Consequently, this problem prompted researchers to introduce honey encryption (HE). HE helps to withstand the vulnerability of the encryption algorithms in brute force attacks making transmitted data more secure and efficient. Hence to reduce processing time we proposed a hybridized HE with Residue Number System. Traditional Moduli Set {2n-1, 2n, 2n+1} was used to generate the Key while Chinese Remainder Theorem (CRT), and HE technique were used for decrypting the data, based on the Distribution Transformation Decoder. Thus, anytime an attacker tries to access the data, so long the length of the guessed key is the same as the original key, the HE algorithm will generate meaningful but fake data, this helps to deter an attacker from further threat. The result showed that the proposed system was able to withstand brute force attacks with less processing time compared to other systems. N-values for the moduli set used were varied against encryption and decryption time. Comparing the obtained results with the existing system, the proposed system processing time was faster and more secure.
A Machine Learning Approach to Dropout Early Warning System Modeling
(International Journal of Advanced Studies in Computer Science and Engineering (Ijascse.）, 2019-02-28) Isiaka, Mope Rafiu; Babatunde, Ronke Seyi; Abdulsalam, Olaniyi Sulaiman
Dropout has been identified as a social problem with an immediate and long term consequences on the socio economic development of the society. A major element of a conceptual framework for dropout monitoring and control system is the prevention component. The component specified the need for automatic dropout early warning system. This paper presents the procedure for building the adaptive model, that is the core element for the realization of the prevention component of the framework. The model development is guided by the knowledge of the domain experts. An experimental approach was used to identify the K-Nearest Neighbors (KNN) as the best of the six algorithms considered for the adaptive models. Other algorithms explored are Logistic Regression (LR), Linear Discriminant Analysis (LDA), Classification and Regression Trees (CART), Gaussian Naive Bayes (NB) and Support Vector Machines (SVM). Historical dataset from a State University in the North Central region of Nigeria was used for building and testing the adaptive model. The average performance of the model on full attributes and the first year academic attributes compared favorably with 0.98 Precision, Recall and F1-Scores. Conversely, the Precision, Recall and F1-Scores on the entry attributes are 0.90, 0.95 and 0.92 respectively. The implication of this is that the academic performance is a very significant factor for dropping out of the educational system. The developed model can provide an early signal on the students with the propensity to dropout, thereby serving as an advisory system for the students, parents and school management towards curtailing the menace.
A Neuro-Fuzzy Model for E-Learning Activities
(LAMBERT Academic Publishing, 2017) Isiaka, Mope Rafiu
E-learning system is an Information Technology-based platform adopted globally to complete or supplement the tradition mode of education. Most existing works e-learning system conceptualized on Learning Technology System Architecture (LTSA) Standard for content adaptability and interoperability but played less on instructional design theory and dynamic learners' activities monitoring. This study therefore extends LTSA with Instructional Design (ID) subsystem and a neutro-fuzzy artificial intelligent model for automatic interpretation of learner's activities. The learning resource store in the LTSA was extended with a process for five selected instructional design theories (Gagnes, Dick and Carey, ADDIES, Merrills and Kemp). The extended LTSA was implemented on open-sourced Modular Object Oriented Distance Learning Environment (Moodle). The Students' Impression (SI) on the extended LTSA was investigated through Survey Monkey online platform using sixty (60) 300 level, B.Sc Computer Science students that offered System Analysis and Design (CSC 306) during the 2013/2014 session at Kwara State University, Malete as a case study. The effectiveness of the extended LTSA for the students under consideration was determined on the basis of mean score of the class examination grades. Furthermore, the evaluation process in the LTSA was extended to include diagnostic process, after which a six layer feed-forward neuro-fuzzy network model for learners' activities was developed and integrated. The developed neuro-fuzzy model was simulated in Matrix Laboratory editor. The performance of the developed model was evaluated by comparing the developed model with the existing Mamdani Fuzzy, Sugeno Fuzzy and Sigmoid neuro-fuzzy learner' activities models in terms of Mean Absolute Error (MAE) and Root Mean-Squared Error (RMSE). The complexity of the developed model was determined on the basis of accuracy, True Positive (TP), False Positive (FP), True Negative (TN) and False Negative (FN). The effect of the extended LTSA on the academic performance of the students vielded 56.98 class mean score. The MAE and RMSE obtained for the developed model are 127.4 and 11.3 respectively while those of Mamdani Fuzzy are 420.9 and 20.6 respectively. Also, these values for Sugeno Fuzzy are 2190.5 and 46.9 respectively while those of Sigmoid neuro-fuzzy are 157.6 and 12.6 respectively. Furthermore, complexity of the developed model assumed an accuracy value of 0.9, TP value of 67.0. FP value of 9.0, TN value of 74.0 and FN of 0.0. The Triangle model which shares the same accuracy value of 0.9 with the developed model has TP value of 66.0, FP value of 18.0, TN value of 65.0 and FN of 1.0. Consequently, the integrated ID-subsystem facilitates grounded instruction development that provides improved academic performance for the class. Also, the developed neuro-fuzzy model which outperformed the other three models on the complexity metrics offers seamless evaluation of learners' activities.
Agentic RAG-Driven Multi-Omics Analysis for PI3K/AKT Pathway Deregulation in Precision Medicine
(Multidisciplinary Digital Publishing Institute (MDPI), 2025-08-30) Arowolo, Olaolu Micheal; Arowolo, Olaolu Micheal; Isiaka, Mope Rafiu; Igulu, Kingsley Theophilus; Balogun, Bukola Fatimah; Popescu, Mihail; Xu, Dong
The phosphoinositide 3-kinase (PI3K)/AKT signaling pathway is a crucial regulator of cellular metabolism, proliferation, and survival. It is frequently dysregulated in metabolic, cardiovascular, and neoplastic disorders. Despite the advancements in multi-omics technology, existing methods often fail to provide real-time, pathway-specific insights for precision medicine and drug repurposing. We offer Agentic RAG-Driven Multi-Omics Analysis (ARMOA), an autonomous, hypothesis-driven system that integrates retrieval-augmented generation (RAG), large language models (LLMs), and agentic AI to thoroughly analyze genomic, transcriptomic, proteomic, and metabolomic data. Through the use of graph neural networks (GNNs) to model complex interactions within the PI3K/AKT pathway, ARMOA enables the discovery of novel biomarkers, probable candidates for drug repurposing, and customized therapy responses to address the complexities of PI3K/AKT dysregulation in disease states. ARMOA dynamically gathers and synthesizes knowledge from multiple sources, including KEGG, TCGA, and DrugBank, to guarantee context-aware insights. Through adaptive reasoning, it gradually enhances predictions, achieving 91% accuracy in external testing and 92% accuracy in cross-validation. Case studies in breast cancer and type 2 diabetes demonstrate that ARMOA can identify synergistic drug combinations with high clinical relevance and predict therapeutic outcomes specific to each patient. The framework’s interpretability and scalability are greatly enhanced by its use of multi-omics data fusion and real-time hypothesis creation. ARMOA provides a cutting-edge example for precision medicine by integrating multi-omics data, clinical judgment, and AI agents. Its ability to provide valuable insights on its own makes it a powerful tool for advancing biomedical research and treatment development.
An Enhanced Learning Technology System Architecture for Web-Based Instructional Design
(International Journal of Emerging Technologies in Learning (iJET), 2016-02-01) Isiaka, Mope Rafiu; Omidiora, Olusayo Elijah; Olabiyisi, Olatunde Stephen; Okediran, O. Oludotun
Instructional design (ID) models are proven prescriptive techniques for qualitative lessons that could guarantee learning. Existing Learning Management Systems (LMS) miss-out the roles of this important quality control mechanism by providing a mere plane and passive platform for content authoring, thus becomes vulnerable for poor instructional design. This paper demonstrates an effort to ameliorate this limitation by extending the IEEE Learning Technology System Architecture (LTSA) with ID design processes. The Use Case diagram, Activities diagram and Entity Relation diagram for the extended LTSA are presented. The extended architecture was implemented on Moodle open sourced LMS which was extended and hosted live. Students’ impressions on the functionalities and operational effects of the platform were collated using online survey. The academic effects of the platform on the students’ performance were determined using the class mean. The value obtained was compared with that of the control group in the same session and those from the previous sessions. Consequently, this work demonstrates the feasibility of integrating ID models in E-learning. It also justifies its effects on the quality of learning.
An Improved Coronary Heart Disease Predictive System Using Random Forest
(Asian Journal of Research in Computer Science, 0021-08-11) Abdulraheem Abdul; Isiaka, Mope Rafiu; Babatunde, Ronke Seyi; Ajao, Jumoke Falilat
Aims: This work aim is to develop an enhanced predictive system for Coronary Heart Disease (CHD). Study Design: Synthetic Minority Oversampling Technique and Random Forest. Methodology: The Framingham heart disease dataset was used, which was collected from a study in Framingham, Massachusetts, the data was cleaned, normalized, rebalanced. Classifiers such as random forest, artificial neural network, naïve bayes, logistic regression, k-nearest neighbor and support vector machine were used for classification. Results: Random Forest outperformed other classifiers with an accuracy of 98%, a sensitivity of 99% and a precision of 95.8%. Feature selection was employed for better classification, but no significant improvement was recorded on the performance of the classifier with feature selection. Train test split also performed better that cross validation. Conclusion: Random Forest is recommended for research in Coronary Heart Disease prediction domain.
Assessing the Adequacy of Artificial Intelligence Tools for African Narratives: Towards Responsible and Culturally Inclusivity
(FUW Trends in Science & Technology Journal, 2025-11-28) Isiaka, Mope Rafiu; Jimoh, Muhammed Kamaldeen; Abdulsalam, Olaniyi Sulaiman; Olanrewaju, Olaitan Toyyib
The increasing reliance on Generative Artificial Intelligence (GenAI) across diverse domains has sparked global attention regarding its adequacy and accuracy. Specifically, its capacity to capture and represent African narratives remains underexplored. Most AI systems are developed within Western contexts fail to align with Africa's diverse sociocultural realities, thereby perpetuating biases, misrepresentation, and the erasure of indigenous knowledge systems. This study critically evaluates the performance of the ChatGPT, DeepSeek, Gemini, and Perplexity large language models in processing and representing African narratives. Using a mixed-methods approach, it incorporates systematic assessments of selected popular AI tools and qualitative input from domain experts in African linguistics, culture and technology. A survey involving academic staff from state and federal universities in Nigeria’s North-Central region contributed 24 relevant prompts, which were combined with 15 research-generated prompts, totalling 39. These prompts were executed concurrently on the selected AI models, and the resulting outputs were evaluated by subject-matter experts for accuracy, adequacy, and credibility. Perplexity consistently achieved the highest ratings across all parameters, whereas the other models displayed varying degrees of effectiveness. Notably, the findings revealed a “white as default” bias and a tendency to prioritise content from Eastern and Southern Africa. The study also identified serious gaps in the handling of African languages, idioms, and culturally embedded expressions, stemming from the poor representation of low-resource languages, limited infrastructure, skill deficiencies, and weak governance. In response, this study proposes roadmaps for responsible AI development tailored to African contexts, advancing ethical practices and amplifying African voices in the digital era.
Beyondthehypesondatarebalancinginimbalancelearning: towards abalancedframeworkandrecommendersystem
(International Journal of Data Science and Analytics, 2025-07-21) Isiaka, Mope Rafiu; Olorede, Opeyemi Kabir; Babatunde, Ronke Seyi; Ajao, Jumoke Falilat
Class-imbalanced learning presents critical challenges in machine learning, largely because data in most domains are naturally imbalanced. Although resampling techniques have been widely applied to address this issue, their effectiveness has been inconsistent and sometimes flawed, owing to artificial assumptions. In this study, we move beyond hype surrounding resampling methods by exploring alternative strategies, such as ensemble learning, cost-sensitive algorithms, and one-class classification techniques. Through rigorous experimentation across extreme, moderate, and mild imbalance levels, our findings reveal that these alternatives often outperform traditional resampling in terms of F1-scores, with ensemble SVM and one class logistic regression achieving notable values of 1.00 and 0.90, respectively. In addition, we introduce a knowledge-based recommender system designed to help practitioners choose the most appropriate techniques for addressing class imbalance. This research argues that resampling is not always the optimal solution for all instances, thereby advocating a more balanced framework that leverages advanced methods for superior performance in imbalanced learning tasks. Our study advances the field by offering a pragmatic, data-driven approach to overcoming class imbalances, contributing valuable insights for both researchers and practitioners.
Development of a Language Translator Model for Yoruba Language Using Bidirectional Long Short-term Memory Encoder
(KWASU Journal of Information, Communication and Technology (KJICT.), 2024-11-01) Ajao, Jumoke Falilat; Ajao, Olorundare Abdulazeez; Isiaka, Mope Rafiu; Yusuff, Ronke Shakirat; Yahaya, Abdullahi
Nigeria is a nation of vast ethnic diversity, encompassing a multitude of cultures, languages, and values. These differences often present significant communication barriers among its various ethnic groups. To bring the language barrier, Google has integrated several Nigerian languages into its translation model. This study focuses on the development of an automatic speech recognition (ASR) system for Yoruba, one of Nigeria’s indigenous languages. Speech dataset was collected from native Yoruba speakers and subjected to extensive preprocessing to eliminate background noise introduced during the recording process. The preprocessed data was then transformed into a speech spectrogram, serving as input to a bidirectional long short-term memory (BiLSTM) model. The hyperparameters of the BiLSTM model were optimised for improved performance. The translation model was evaluated using metrics such as BLEU, METEOR, and ROUGUE. Results demonstrated significant improvements in the ASR model’s ability to accurately transcribe and translate Yoruba language speech. This advancement in translation accuracy highlights the potential for better cross-linguistic communication among Nigeria’s ethnic groups, fostering greater inclusivity and understanding. This study contributes to ongoing efforts to incorporate underrepresented languages in global translation models, addressing the challenge of language diversity in multilingual societies.
Development of a Web-Based Platform Utilizing a Hybrid CNN Architecture for Real-Time Fecal Image Classification
(Ajayi Crowther Journal of Pure and Applied Sciences, 2025-10-29) Omideyi, Damilare Andrew; Isiaka, Mope Rafiu; Babatunde, Ronke Seyi
Accurate classification of fecal images offers a non-invasive and scalable approach to detecting enteric diseases such as Coccidiosis, Salmonellosis, and Newcastle disease. Traditional diagnostic methods remain limited by their dependence on expert analysis, time consumption, and cost. This study presents the development of a web-based platform utilizing a hybrid Convolutional Neural Network (CNN) architecture for the real-time classification of fecal images. The proposed model combines MobileNetV2 for lightweight inference and VGG-16 for deep feature extraction, ensuring both accuracy and computational efficiency. Following training on a curated and augmented dataset of labeled fecal images, the model was quantized and converted into TensorFlow Lite format for mobile compatibility. The final system enables users to upload images via a browser or Android application and receive instant classification into four categories: Coccidiosis, Salmonella, Newcastle disease, or Healthy. Performance evaluations confirm high diagnostic accuracy and responsiveness under real-world conditions. This study demonstrates that hybrid deep learning models, when integrated into web-based platforms, can support effective classification of fecal images for field-ready disease detection. It is recommended that future work expands the dataset across multiple geographic regions and incorporates additional diagnostic inputs to enhance system robustness and adaptability.
Development of an Intrusion Detection System in Web Applications Using C Means and Decision Tree Algorithm
(International Journal of Applied Science and Technology, 2023-03-05) Isiaka, Mope Rafiu; Popoola, Damilola David
Intrusion detection is extremely important for online applications and for determining whether there has been a hostile entrance into the website. The aim of this research is to provide a machine learning technique for detecting intrusion in a web application. Machine learning models such as C-means, Decision Tree and Support Vector Machine were utilized to create an intrusion detection system. The study used the CIC-IDS 2018 intrusion dataset (Friday-Working Hours-Afternoon Ddos.pcap ISCX). The data was initially sent to Decision tree and SVM which had accuracy of 99.97% and 99.77%, respectively. The raw data was next transferred into the c-means clustering approach, which had an accuracy of 99.99%. The goal of the clustering technique used is to improve the system’s accuracy, and the results were assessed using performance metrics like accuracy, sensitivity, precision, specificity, F1-score as well as accuracy comparison of the results obtained with the state of the art.
High-speed Reverse Converter (Hisprec) for Improving Reliability in Wireless Sensor Networks
(Technoscience Journal for Community Development in Africa, 2020-06-15) Raji, Ayodele Kamaldeen; Yusuf-Asaju, Wuraola Ayisat; Isiaka, Mope Rafiu
Wireless Sensor Networks (WSNs) consist of huge quantities of sensor hubs that are distributed in an unfriendly environment which are used to monitor the environment and check physical conditions. It has been proven that vitality, speed and reliability are the most obliging components on the functionality of WSNs as they are controlled with constrained vitality and limited hardware resources. Thus, WSN must be designed in an energy-efficient and reliable manner. However, the intrinsic properties of carry-free operations, parallelism and adaptation to internal failure of Residue Number System (RNS) have made it a promising strategy for high-speed arithmetic and low power utilisation for WSN applications. Hence, to reduce power utilisation and improve message reliability in WSNs, HIgh SPeed REverse Converter (HISPREC) is proposed in this article. The paper is aimed at designing high speed and low vitality utilisation reverse converter to improve the reliability and diminish conversion delay in real-time WSNs. The proposed reverse converter (HISPREC) is designed for the moduli set {2n+1-1, 2n, 2n-1} and actualised using Mixed Radix Conversion (MRC). The results of theoretical and experimental simulations show that HISPREC outstripped existing reverse converters as far as speed, hardware requirements, vitality utilisation and conversion delay all of which affect the reliability of the received information in real-time WSNs.
Image Restoration: A LSHADE-GAN Approach for Yoruba Documents
(Proceedings of International Computing and Communication Conference (13C 2024), 2024-04-22) Lawal, Olamilekan Lawal; Ajao, Jumoke Falilat; Isiaka, Mope Rafiu
The paper introduces an innovative approach to reconstructing historical handwritten texts, specifically focusing on Yoruba documents. This method combines a generative adversarial network (GAN) with the LSHADE algorithm to create the LSHADE-GAN model. Trained on a curated dataset of five degraded Yoruba documents, this model surpasses traditional image-processing techniques and deep learning-based methods in performance. Evaluation of the LSHADE-GAN model using two samples reveals F-measures of 61.83% and 78.02%, showcasing its superiority over DE-GAN (58.24% and 75.23%) and PSO-GAN (51.67% and 66.46%) approaches. Additionally, the model demonstrates enhanced PSNR and visual quality, underscoring its effectiveness in preserving cultural heritage through accurate reconstruction of historical texts.
Implication of Variation in Ant Colony Optimization Algorithm (ACO) Parameters on Feature Subset Size
(Computing and Information Systems Journal, 2017) Babatunde, Ronke Seyi; Isiaka, Mope Rafiu; Oyeranmi J. Adigun; Babatunde, N. Akinbowale
Purpose: The purpose of this research is to apply ant colony optimization (ACO), a nature inspired computational (NIC) technique to achieve optimal feature subset selection, for the purpose of data dimensionality reduction, which will remove redundancy, provide a reduced storage space, improve memory utilization and subsequently, classification task. Design/Methodology/Approach: The ACO feature subset selection process was filter-based, accomplished by using correlation coefficient as the heuristic information which guides the search to determine the pixel (feature) attractiveness. The intensity of the pixel represents the pheromone level . Also, ρ, φ, and are key parameters which were tuned to various arbitrary values at each iteration to determine the best combination of parameters which can result in a reduced feature subset (window size). The procedure was simulated in MATLAB 2012. Findings: The study has been able to demonstrate that experimental results obtained by tuning the parameters of ACO at each run gave a reduced feature subset (window size), at parameter settings of α (0.10), β (0.45), ρ (0.50), φ (1.0) which was salient for efficient face recognition thereby demonstrating the effectiveness and superiority of filter-based over wrapper–based feature selection. Research Limitation: The major limiting factor in this research was time and materials which hindered the further testing and implementation of innovative computational intelligence. Originality/Value: This research work is valuable because it gives the implication of the variation of the ACO parameters which determines the extent to which the optimal solution construction that results in a reduced feature subset can be realized.
Implications of System Identification Techniques on ANFIS E-learners Activities Models-A Comparative Study
(Transactional on Machine Learning and Artificial Intelligence, 2016-02-21) Isiaka, Mope Rafiu; Omidiora Elijah Olusayo; Olabiyisi Stephen O.; Okediran Oladotun O.; Babatunde Ronke Seyi
Efficient e-learners activities model is essential for real time identifications and adaptive responses. Determining the most effective Neuro- Fuzzy model amidst plethora of techniques for structure and parameter identifications is a challenge. This paper illustrates the implication of system identification techniques on the performance of Adaptive Network based Fuzzy Inference System (ANFIS) E-learners Activities models. Expert knowledge and Historical data were used to formulate the system and their performances were compared. Similarly, comparison was made between memberships functions selected for Historical data identification. The efficiencies of the simulated models in MATLAB editor were determined using both classification uncertainty metrics and confusion matrix–based metrics. The classification uncertainty metrics considered are Mean Absolute Error (MAE) and Root Mean-Squared Error (RMSE). The confusion matrix-based metrics used are Accuracy, Precision and Recall. It was discovered that the model based on Experts Knowledge after training outperformed those based on Historical Data. The performances of the Membership Functions after ranking are Sigmoid, Gaussian, Triangular and G-Bells respectively.
Improved Cloud-Based N-Primes Model for Symmetric-Based Fully Homomorphic Encryption Using Residue Number System
(Springer Nature Switzerland, 2021) Jimoh, Muhammed Kamaldeen; Isiaka, Mope Rafiu; A. W. Asaju-Gbolagade; K. S. Adewole; K. A. Gbolagade
Encryption schemes that allow computation to be performed on an encrypted data are required in modern real-world applications. This technology is a necessity for cloud computing, processing resources and share storage, in order to preserve integrity and privacy of data. The existing partial and fully homomorphic encryption (PHE and FHE) for both asymmetric and symmetric approaches are still suffering from efficiency in terms of encryption execution time and very large ciphertext file produced at the end of encryption. In this paper, we consider symmetric approaches and focus on overcoming the drawbacks of N-prime model using residue number system (RNS). The experimental results obtained show that the proposed RNS-based N-prime model for symmetric-based FHE improves the system latency and reduces the ciphertext file expansion by approximately 72% as compared to the existing N-prime model. The proposed improved N-prime model is guaranteed to provide optimum performance and reliable solution for securing integrity and privacy of user’s data in the cloud.

Browsing by Author "Isiaka, Mope Rafiu"

Results Per Page

Sort Options