Browsing by Author "Ajao, Jumoke Falilat"

Now showing 1 - 7 of 7
  • Thumbnail Image
    Item
    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.
  • Thumbnail Image
    Item
    A Study of Impacts of Artificial Intelligence on COVID-19 Prediction, Diagnosis, Treatment, and Prognosis
    (Journal of Advances in Mathematical & Computational Sciences, 2022-12-04) Isiaka, Rafiu Mope; Babatunde, Ronke Seyi; Ajao, Jumoke Falilat; Yusuff, Shakirat Ronke; Popoola, Damilola David; Arowolo, Michael Olaolu; Adewole, Kayode
    Following the identification of Coronavirus Disease 2019 (COVID-19) in Wuhan, China in December 2019, AI researchers have teamed up with a health specialist to combat the virus. This study explores the medical and non-medical areas of COVID-19 that AI has impacted: the prevalence of the AI technologies adopted across all stages of the pandemic, the collaboration networks of global AI researchers, and the open issues. 21,219 papers from ACM Digital, Science Direct and Google Scholar were examined. Adherence to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework and utilizing the PICO (population, intervention, comparison, and outcome) paradigm, guided the inclusion of researches in the review. Tables and graphs were utilized to display the results. Analysis revealed that AI has impacted 4 molecular, 4 clinical, and 7 societal areas of COVID-19. Deep Learning among other AI technologies was traced to all aspects of the pandemic. 2173 authors and co-authors were traced to these achievements, while 32 of the most connected 51 authors were affiliated with institutions in China, 18 to the United States, and 1 to Europe. The open issues identified had to do with the quality of datasets, AI model deployment, and privacy issues. This study demonstrates how AI may be utilized for COVID-19 diagnosis, prediction, medication and vaccine identification, prognosis, and contact person monitoring. This investigation began at the beginning of the epidemic and continued until the first batch of vaccinations received approval. The study provided collaboration opportunities for AI researchers and revealed open issues that will spike further research toward preparing the world for any future pandemic.
  • Thumbnail Image
    Item
    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.
  • Thumbnail Image
    Item
    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.
  • Thumbnail Image
    Item
    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.
  • Thumbnail Image
    Item
    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.
  • Thumbnail Image
    Item
    Use of White Shark Optimization for Improving the Performance of Convolution Neural Network in Classification of Infected Citrus
    (University of Ibadan Journal of Science and Logics in ICT Research, 2023-06-09) Babatunde, Ronke Seyi; Isiaka, Mope Isiaka; Ajao, Jumoke Falilat; Abdulsalam, Olaniyi Sulaiman; Balogun, Bukola Fatimah
    Citrus plant diseases are major causes of reduction in the production of citrus fruits and their usage. Early detection of the onset of the diseases is very important to curb and reduce its spread. A number of researches have been done on the detection and classification of the diseases, most of which have identified poor labelling of symptoms which result into improper classification. Some researchers have also experimented on the effectiveness of convolution neural network and other deep learning techniques, most of which results into a faster convergence but suffers from low accuracy, computational overhead and overfitting as fundamental issues. To reduce the effects of overfitting, this research developed a White Shark Optimization-Convolution Neural Network (WSO-CNN) technique to address the aforementioned problem by introducing a regularization strategy via feature selection which selects more useful and distinguishing features for classification. As a result, the developed technique was able to detect and classify various types of citrus fruit diseases and label them accordingly with low false positive rate, high sensitivity, specificity, increased accuracy and reduced recognition time, based on all the experiments performed with the dataset used in the research. Hence WSOCNN performed better than CNN in classifying citrus plant disease having a reduced FPR of 3.57%, 8.34%, 3.89% and 9.00% for black spot, greasy spot, canker and healthy/non healthy dataset respectively.