Browsing by Author "Oloyede, Isiaka"
Now showing 1 - 11 of 11
Results Per Page
Sort Options
Item BAYESIAN CLASSIFICATION OF HIGH DIMENSIONAL DATA WITH GAUSSIAN PROCESS USING DIFFERENT KERNELS(Anale. Seria Informatică. Annals. Computer Science Series., 2018) Oloyede, IsiakaThe study investigates asymptotic classification of high dimensional data by adopting Gaussian Process, five different kernels(covariance functions) were employed and compared to showcase the outperformed kernel asymptotically. Log marginal likelihood, Accuracy and log loss were the measurement criteria adopted to measure classification performances. The study therefore observed that the classification performed well asymptotically and found out that Gaussian Process Maximum Likelihood(GPML) had overall best model improvement asymptotically and across the covariance structures. K3 and K4 had the best accuracy in classification paradigm at the lower sample sizes but GPML and learned kernel had best model accuracy as the sample sizes tend to large sizes.Item BAYESIAN HETERO-LASSO (A GIBBS SAMPLING APPROACH)(Journal of Science, Technology, Mathematics and Education (JOSTMED), Federal University of Technology, Minna, 2018-12) Oloyede, IsiakaThe study investigates the asymptotic consistency and efficiency of Bayesian estimator due to violation of homoscedasticity cum non-multicollinearity properties. Mean Square Error( MSE) and Bias were the performance measuring criteria on twin non-spherical disturbances. The seed was to 12345; were set at = 2.5,1.5,1,0,0,0.5 ; Xs variables as design matrix was generated from the multivariate normal distribution with > 0 and . and were contaminated with Harvey (1976) heteroscedastic error structure; ,… were collinear covariate with pairwise correlation of 0.9, the sample sizes were set as 25, 50,70,100,200,500 and 1000. The number of replications of the experiment was set at 11,000 with burn-in of 1000 which specified the draws that were discarded to remove the effect of the initial values. The thinning was set at 5 to ensure the removal of the effect of autocorrelation in our MCMC simulation. In this paper, the study was able to depict the asymptotic consistency and efficiency of the hetero-lasso estimator at large sample sizes, the study affirmed that Bayesian hetero-lasso estimator performed well when the sample size is large. The outcome of the study revealed improved performances of the estimator in the model parameter estimates asymptotically.Item Bayesian Hierarchical Modeling of Asymmetric Effect of Autocorrelated Error(Al-Hikmah Journal of Pure & Applied Sciences, Al-Hikma University, Ilorin., 2015) Oloyede, IsiakaThis study investigates the asymmetric effect of autocorrelated error of hierarchical model via Bayesian paradigm. The study employed full Bayesian experiment by considering the marginal conditional posteriors density of the model parameters estimate. The extreme cases of autocorrelated error were considered by selecting -0.99 and 0.99 for rho. The seed was set to 12345; were set at 2.5, 1.5, 0.5; Xs variables were generated using uniform distribution. The number of replications of our experiment was set at 11,000 with burn-in of 1000 which specified the draws that were discarded to remove the effect of the initial values. The thinning was set at 5 to ensure removal of the effect of autocorrelation in Markov Chain Monte Carlo simulation. The study revealed that positive correlation had higher impact than negative correlation when the magnitude is 0.9; whereas at lower correlation, negative correlation had higher impact. The study affirmed improvement in consistency and efficiency on the model parameters estimates.Item Bayesian Minimum Message length87 with parametric heteroscedasticity model(Nigerian association of Mathematical Physics, 2015) Oloyede, IsiakaA Metropolis Hasting algorithm was adopted to perform simulation on marginal posterior distribution of heteroscedastic linear model using minimum message length87 which was conjugated with normal and inverted gamma priors to derive joint posterior distributions. The asymptotic behaviour was compared using absolute bias and mean square error criteria in order to ascertain consistency and efficiency of the estimator. The estimator is both asymptotically consistent and efficient.Item A Bayesian Network for Predicting Body Composition of Human Body(Centrepoint Journal (Science Edition), university of ilorin, 2016) Oloyede, IsiakaThis study investigates the predictive body composition adopting age, height, weight, waist circumference and body mass index as the covariates to predict three components that form the body weight: bone, fat and lean. Both saturated Bayesian network and convenient Bayesian network were compared using absolute bias, standard deviation and root mean square error to ascertain consistency and efficiency of the two types of Bayesian networks. The Directed Acyclic Graph were used to compare the two networks; nine different samples sizes were constructed using uniform distribution for both dependent variables and covariates. The study observed that saturated Bayesian network outperformed convenient Bayesian network both at small and medium sample sizes whereas the two networks estimators converged at higher sample sizes yielding the same standard deviation and root mean squares error. The study therefore recommends that saturated Bayesian network should be used at small sample sizes while any of the two types could be used at higher sample sizes.Item Bayesian Spatial heteroscedastic autoregressive model(Nigerian association of Mathematical Physics, 2018) Oloyede, Isiakathe study investigates the impact of spatial heteroscedastic autoregressive model on parameter estimations by adopting white(1980) heteroscedastic covariance matrix Bayesian paradigm. Coordinate references of 94 cities in Nigeria were collected, the study obtained the minimum and maximum values and these were used to generate coordinate references for the entire simulated data using uniform distribution.Item BOOTS TRAPPING SUPERVISED CLASIFIER PARADIGM(Anale. Seria Informatică. . Computer Science Series., 2017) Oloyede, IsiakaThe study investigates the classification of learning algorithms in a bootstrap paradigm, the study examined features classification with binary class attributes in a bootstrap paradigm. Support Vector Machine, k-Nearest Neighbour, Random Forest, rpart, Artificial Neural Network and Naïve Bayes learning algorithms were compared. Accuracy, Prediction error, Sensitivity and Specificity were used as assessment criteria of the classifier after tuning to have minimum cost. The study therefore sample the training set and classifying each of the training set, the summary of the prediction error was obtained based on the testing dataset, the study showed that artificial neural network outperformed other learning algorithms with respect to accuracy criterion whereas the celebrated support vector machine performed poorly amongst the learning algorithms considered, the study depicted that artificial neural network outperformed other learning algorithms with the least misclassification error. The study depicted that K nearest neighbour outperformed other learning algorithms with highest sensitivity while ANN outperformed other learning algorithms with highest specificity. This study affirmed that there would be need to use more than a learning algorithm when there are irrelevant features in the data sets.Item comparison of mixture experiments models(Journal of Information , education , science and technology, federal university of technology, minna., 2018) Oloyede, IsiakaThe study investigates various mixture experiment model and proposed a new polynomial model which is a modified special cubic model. The models were examined via bootstrap resampling scheme and select model based on bias , aic, bic and standard error criteria. The study observed that the modified model was the second best model.Item comparison of some estimators of bayesian heteroscedastic linear model(ABACUS, Mathematical Association of Nigeria, 2016) Oloyede, IsiakaIn order to investigate the asymptotic consistency and efficiency of estimators with normal-gamma double sided heteroscedastic error structure, the study explored full bayesian metropolis hasting , algorithm experiments, an approach of markov chain monte carlo simulation. The study contaminated the model with one component of two sided error strucuture. A metropolis hasting adopted to perform simulation on the marginal posterior distribution of heteroscedastic linear econometric model. Absolute bias and mean squares error criteria were used to evaluate finite properties of the estimators.Item E ciency of bayesian heteroscedastic linear model(Electronic Journal of Applied Statistical Analysis EJASA, Electron. J. App. Stat. Anal., 2014) Oloyede, Isiaka; Ipinyomi, R.A; Iyaniwura, J.O.In order to investigate the asymptotic e ciency of estimators under two di erent simulation techniques, normal-normal double sided Heteroscedas- tic error structure was adopted. We explored Direct Monte Carlo method of Zellner et al. (2010) and Metropolis Hasting Algorithm experiments, an approach of Markov Chain Monte Carlo. We truncated the model with one error component of two sided error struc- ture. A Metropolis-Hasting Algorithm and Direct Monte Carlo adopted to perform simulation on marginal posterior distribution of heteroscedastic lin- ear econometric model. Since Ordinary Least squares is invalid and inef- cient in the presence of heteroscedastic, heteroscedastic linear model was conjugated with informative priors to form posterior distribution. Maximum Likelihood Estimation was compared with Bayesian Maximum Likelihood Estimation, Mean Squares Error criterion was use to identify which esti- mator and/or simulation method outperform other. We chose the following sample sizes: 25; 50; 100; and 200. Thus 10,000 simulations with varying degree of heteroscedastic error structures were adopted. This is subject to the level of convergence. In the overall, minimum mean squares error criterion revealed improving performance asymptotically regardless of the degree of heteroscedasticity. The results showed that Direct Monte Carlo Method outperformed Markov Chain Monte Carlo Method and Maximum Likelihood Estimator with mini- mum mean square error at any degree of heteroscedasticity.Item SEMI BAYESIAN INFERENCE HIGH AND LOW DIMENSIONAL DATA WITH MULTICOLLINEARITY(Journal of Science, Technology, Mathematics and Education (JOSTMED), Federal University of Technology, Minna, 2018-03) Oloyede, IsiakaIt is generally known that correlation amongst features in high and low dimensional data lead to parameters that artificially insignificance. This study investigates asymptotic properties of some semi-bayesian estimators and compared it with non-bayesian estimator in the presence of multicollinearity. Variational and Empirical Bayes estimators were succinctly compared with ordinary least squares estimator using bias, mean squares error (MSE) and predictive mean squares error (PMSE). The number of iteration was 1000. In high dimensional data, it was found that empirical Bayes Linear Regression (EBLR) outperformed other estimators whereas OLS performed poorly using the PMSE as evaluation criterion. The study found out that in low dimensional data, variational Bayes Linear Regression (VBLR) outperformed other estimators yet OLS performed poorly using the PMSE criterion. Asymptotically, the three estimators were inconsistent but having the same pattern in low dimensional data but they were fairly consistent between the sample sizes 30 to 50 using the bias criterion. The study therefore concluded that empirical Bayes estimator should be adopted in high dimensional data while variational Bayes should be adopted in low dimensional data.