Browsing by Author "Oyebayo Ridwan Olaniran"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    A Novel Variable Selection Procedure for Binary Logistic Regression Using Akaike Information Criteria Testing: An Example in Breast Cancer Prediction.
    (Turkiye klinikleri, 2023-07-13) Oyebayo Ridwan Olaniran; Saidat Fehintola Olaniran
    Breast cancer is a leading cause of cancer-related death among women worldwide, with approximately 2.3 million new cases and 685,000 deaths reported in 2020 alone. One critical step in developing effective classification and prediction models is variable selection, which involves identifying a subset of relevant variables from a larger set of potential predictors. Accurate variable selection is crucial for building interpretable and robust models that are not overfit to noise, leading to improved model performance and generalization ability. In this paper, we proposed an alternative objective approach for comparing two Akaike Information Criterions (AIC) that originated from two competing models, such that the magnitude of the difference is subjected to the statistical test of significance. Material and Methods: We developed a new backward elimination variable selection procedure similar in spirit to the existing “step AIC” within the environment of R statistical software. We used both simulated and Wisconsin breast cancer diagnostic datasets to compare the proposed method's variable selection and predictive performances with “step AIC” and LASSO. Results: The simulation showed that the proposed AIC procedure achieved higher variable selection sensitivity, specificity and accu racy when compared to stepAIC and LASSO. Also, the proposed AIC method's prediction results are relatively comparable with ste pAIC and LASSO at various simulated data dimensions. Similar supremacy results were observed with the breast cancer dataset used. Conclusion: The AIC-based variable selection approach pro posed is a promising method that integrates AIC with statistical testing for improved variable selection in breast cancer classifica tion and predictio
  • Thumbnail Image
    Item
    Bayesian Additive Regression Trees for Predicting Colon Cancer: Methodological Study (Validity Study).
    (Turkiye klinikleri, 2022-05-01) Oyebayo Ridwan Olaniran; Saidat Fehintola Olaniran; Jumoke Popoola; Omekam Ifeyinwa Vivian
    The occurrence of colon cancer starts in the inner wall of the large intestine. The survival of colon cancer patients strongly relies on early detection. Diagnosing colon cancer using clinical approaches often takes longer, especially in most developing countries with limited facilities. The recent use of microarray technology has presented a new approach for the oncologist to diagnose cancer cells using non-clinical machine learning methods. In this paper, the aim is to predict the status of colon cancer tissues using the Bayesian Additive Regression Trees (BART) and 2 other machine learning methods. Material and Methods: The development and comparative analysis of BART alongside 2 other competing methods (Random Forest: RF and Gradient Boosting Machine: GBM) were implemented. The dataset used for the analysis is the microarray colon cancer data which consists of 2,000 gene expression …
  • Thumbnail Image
    Item
    Bayesian Regularized Neural Network for Forecasting Naira-USD Exchange Rate
    (Springer, 2022-05-04) Oyebayo Ridwan Olaniran; Saidat Fehintola Olaniran; Jumoke Popoola
    The Artificial Neural Network Autoregressive model (ANN-AR) is a recently adopted approach for forecasting Naira to the USD exchange rate. The Bayesian Regularized Neural Network (BRNN) is an alternative to ANN-AR based on the probabilistic interpretation of network weights. It is useful for solving overfitting problems inherent in ANN when large historical data are not available. In this paper, we developed a BRNN for modelling the monthly time series data of Naira to USD for four years. Performance analysis was observed using actual exchange rate values for the Year 2017 against the model predicted outcomes based on four evaluation metrics. Results from the analysis established the appropriateness of a BRNN for modelling short-term exchange rates in Nigeria.
  • Thumbnail Image
    Item
    Empirical bayesian binary classification forests using bootstrap prior
    (Science Publishing Corporation-uthm.edu.my, 2018-11-30) Oyebayo Ridwan Olaniran; Mohd Asrul Affendi; Gopal Pillay, Khuneswar; Saidat Fehintola Olaniran
    In this paper, we present a new method called Empirical Bayesian Random Forest (EBRF) for binary classification problem. The prior ingredient for the method was obtained using the bootstrap prior technique. EBRF addresses explicitly low accuracy problem in Random Forest (RF) classifier when the number of relevant input variables is relatively lower compared to the total number of input variables. The improvement was achieved by replacing the arbitrary subsample variable size with empirical Bayesian estimate. An illustration of the proposed, and existing methods was performed using five high-dimensional microarray datasets that emanated from colon, breast, lymphoma and Central Nervous System (CNS) cancer tumors. Results from the data analysis revealed that EBRF provides reasonably higher accuracy, sensitivity, specificity and Area Under Receiver Operating Characteristics Curve (AUC) than RF in most of the datasets used.
  • Thumbnail Image
    Item
    IMPROVED BAYESIAN FEATURE SELECTION AND CLASSIFICATION METHODS USING BOOTSTRAP PRIOR TECHNIQUES
    (Faculty of Computer and Applied Computer Science, Tibiscus University of Timisoara, Romania, 2016) Oyebayo Ridwan Olaniran; Saidat Fehintola Olaniran; Yahya, Waheed Babatunde; Banjoko, Alabi; Garba, Mohammed Kabir; Amusa, Lateef; Gatta F.N
    In this paper, the behavior of feature selection algorithms using the traditional t-test, Bayesian t-test using MCMC and Bayesian two sample test using proposed bootstrap prior technique were determined. In addition, we considered some frequentist classification methods like k- Nearest Neighbor (k-NN), Logistic Discriminant (LD), Linear discriminant analysis (LDA), Quadratic discriminant analysis (QDA) and Naïve Bayes when conditional independence assumption is violated. Two new Bayesian classifiers (B-LDA and B-QDA) were developed within the frame work of LDA and QDA using the bootstrap prior technique. The model parameters were estimated using Bayesian approach via the posterior distribution that involves normalizing the prior for the attributes and the likelihood from the sample in a Monte Carlo experiment. The bootstrap prior technique was incorporated into the Normal-Inverse-Wishart natural conjugate prior for the parameters of the multivariate normal distribution where the scale and location parameters were required. All the classifiers were implemented on the simulated data at 90:10 training-test data ratio. The efficiencies of these classifiers were assessed using the misclassification error rate, sensitivity, specificity, positive predictive value, negative predictive value and area under the ROC curve. Results from various analyses established the supremacy of the proposed Bayes classifiers (B-LDA and B-QDA) over the existing frequentists and Naïve Bayes classification methods considered. All these methods including the proposed one were implemented on a published binary response microarray data set to validate the results from the simulation study.