Browsing by Author "Udomboso, C. G."

Now showing 1 - 20 of 28

An adjusted network information criterion for model selection in statistical neural network models
(JMASM, Inc., 2016) Udomboso, C. G.; Amahia, G. N.; Dontwi, I. K.
In this paper, we derived and investigated the Adjusted Network Information Criterion (ANIC) criterion, based on Kullback’s symmetric divergence, which has been designed to be an asymptotically unbiased estimator of the expected Kullback-Leibler information of a fitted model. The ANIC improves model selection in more sample sizes than does the NIC.
Alternative goodness-of-fit test in logistic regression models
(Medwell Journals, 2011) Nja, M. E.; Enang, E. I.; Chukwu, A. U.; Udomboso, C. G.
The Deviance and the Pearson chi-square are two traditional goodness-of-fit tests in generalized linear models for which the logistic model is a special case. The effort involved in the computation of either the Deviance or Pearson chi-square statistic is enormous and this provides a reason for prospecting an alternative goodness-of-fit test in logistic regression models with discrete predictor variables. The Deviance is based on the log likelihood function while the Pearson chi-square derives from the discrepancies between observed and predicted counts. Replacing observed and predicted counts with observed proportions and predicted probabilities, respectively in a cross-classification data arrangement, the standard error of estimate is proposed as an alternative goodness-of-fit test in logistic regression models. The illustrative example returns favourable comparisons with Deviance and the Pearson chi-square statistics.
Application of regression type estimator in double sampling skills to students’ enrollment in Oyo State
(2019-03) Udomboso, C. G.; Akanbi, O. B.; Afolabi, S. A.
This research derived the precision using Regression Estimation technique with the application of secondary data obtained using the number of students enrollment in 2015 (Auxiliary variable “x”) and 2016 (response variable “y”) respectively in secondary schools of Ibadan, Oyo State, Nigeria for the purpose of obtaining average enrollment figures in the selected state in order to know the bright future of secondary schools in Oyo State in general and to establish the empirical comparison of the optimum variances in obtaining the most efficient estimator in order to satisfy the condition; p2≥ 122124 based on the coefficients of Variation for the validity and reliability, the relative efficiency was also determined based on the conditions attached to the supremacy in terms of the estimated mean square error (variance) whereby the regression line does not pass through the origin from the graph of Relative Efficiency (R.E) against Correlation Coefficients (p) that maintain inverse relation. Proper conclusions and recommendations are made based on findings from the analysis in terms of adequate record keeping among the contemporary states within.
ARIMA model and neural network: a comparative study of crime rate modelling
(2015) James, T. O.; Suleiman, S.; Udomboso, C. G.; Babayemi, A. W.
Crime rate is a serious issue that affects everyone in society. It affects the victims, perpetrators, their families the government and even reality of good governance. In this study forecasting of crime rate using autoregression integrated moving average (AR1MA) model was compared with feed forward neural networks. The J multi software was used for analysis of data gotten from State Police Headquarter in Kebbi State from January 2004 to December 2013 and the series was stationary at first difference and ARIMA (0, 1, 1) was obtained as the best model for the series. This was model by Neural Network using SPSS. In the training of the network, the samples were automatically partitioned in to 73.3% of training and 26.7% of testing. The computational result shows that Artificial Neural Network provides better model than ARIMA by having minimum error in the in-sample and out -of- sample in MAE, MSE, and RMSE with 3.84614, 2.00466 and 1.41586 respectively.
Autoregressive distributed lags (ARDL) modelling of the impacts of climate change on rice production in Kebbi State
(Professional Statisticians Society of Nigeria, 2018) James, T. O.; Babayemi, A. W.; Abdulmuahymin, A. S.; Udomboso, C. G.; Bello, M. L.
Autoregress Distributed Lag (ARDL) is an econometric model that determines the long run a d short run association between the Serial (Stationary/ non-stationary as well as reparameterizing them to Error correction model (EMC). Rice cultivation and production is a major source of income for millions of households around the globe especially in Nigeria. It is also a major staple food, but Climate change poses great threat to the stability and sustainability of rice production for sufficient agricultural system, since most Nigeria consumes rice more than other foods and Kebbi state, is one of the major states contributing to the total rice output of the country. Climate change is the major challenge facing rice production. This study therefore, investigates the long-run and short run effect of factors affecting rice production in Kebbi State. 1000 simulations of data were obtained from the data collected between the period of 2005 to 2016 from the state Ministry of Agriculture. The result showed that rainfall has impact both in the long run and short run; 100% increase in rainfall, will tend to give 99.98% increase in rice production in the long-run. However, temperature tends to show insignificant impact on rice production. The result of this paper facilitate understanding for government and agriculturist in the linkages between climate change variables and rice production which can boost and increases the production of rice in Kebbi State.
Bayesian approach to survival modeling of remission duration for acute leukemia
(2019) Akanbi, O. B.; Oladoja, O. M.; Udomboso, C. G.
The problem of analyzing time to event data arises in a number of applied fields like biology and medicine. This study constructs a survival model of remission duration from a clinical trial data using Bayesian approach. Two covariates; drug and remission status, were used to describe the variation in the remission duration using the Weibull proportional hazards model which forms the likelihood function of the regression vector. Using a uniform prior, the summary of the posterior distribution; Weibull regression model of four parameters ( η, µ,β1, β2, was obtained. With Laplace transform, initial estimates of the location and spread of the posterior density of the parameters were obtained. In this present study, data from children with acute leukemia was used. The information from the Laplace transform was used to find a density for the Metropolis random walk algorithm from Markov Chain Monte Carlos simulation to indicate the acceptance rate (24.55%).
Bio-social correlates of intention to use or not to use contraception: The case of Ghana and Nigeria
(Union for African Population Studies, 2015) Udomboso, C. G.; Amoateng, A. Y.; Doegah, P. T.
Based on the 2008 and 2013 Demographic and Health Survey data of Ghana and Nigeria respectively, statistical neural network and logit regression models were used to examine the effects of selected bio-social factors on the intention to use contraception among never married and ever married women in the two countries. The results showed that on the whole, the SNN model identified more biosocial factors affecting the intention to use contraception in the two countries than did the logic model. Educational attainment, exposure to media, and visitation to a health facility affected intention to use contraception significantly and positively in both countries. On the other hand, number of living children, infrequent sexual intercourse, postpartum amenorrhea, opposition to contraception and lack of access to contraceptives negatively affected intention to use contraception. The study findings have underscored the rational nature of the decisions women make in using contraception or not.
Comparative analysis of rainfall prediction using statistical neural network and classical linear regression model
(Medwell Journals, 2011) Udomboso, C. G.; Amahia., G. N.
Different types of models have been used in modeling rainfall. Since 1990s however, interest has shifted from traditional models to ANN in rainfall modeling. Many researchers found out that the ANN performed better than such traditional models. In this study, we compared a traditional linear model and ANN in the modeling of rainfall in Ibadan, Nigeria. Ibadan is a city in West Africa, located in the tropical rainforest zone, using the data obtained from the Nigeria Meteorological (NIMET) station. Three variables were considered in this study rainfall, temperature and humidity. In selecting between the two models, we concentrated on the choice of adjusted R2 (R-2 ), Akaike Information Criterion (AIC) and Schwarz Information Criterion (SIC). Though, the MSE and R2 were also used, it was concluded from results that MSE is not a good choice for model selection. This is due to the nature of the rainfall data (which has wide variations). It was found that the Statistical Neural Network (SNN), generally performed better than the traditional (OLS).
Comparing ANN and ARIMA model in predicting the discharge of River Opeki from 2010 to 2020
(John Willey & Sons Limited, 2018) Fashae, O. A.; Olusola, A. O.; Ndubuisi, I.; Udomboso, C. G.
Many attempts have been made in the recent past to model and forecast streamflow using various techniques with the use of time series techniques proving to be the most common. Time series analysis plays an important role in hydrological research. Traditionally, the class of autoregressive moving average techniques models has been the statistical method most widely used for modelling water discharge, but it has been shown to be deficient in representing nonlinear dynamics inherent in the transformation of runoff data. In contrast, the relatively newly improved and efficient soft computing technique artificial neural networks has the capability to approximate virtually any continuous function up to an arbitrary degree of accuracy, which is not otherwise true of other conventional hydrological techniques. This technique corresponds to human neurological system, which consists of a series of basic computing elements called neurons, which are interconnected together to form networks. The aim of the study is to compare the artificial neural network and autoregressive integrated moving average to model River Opeki discharge (1982–2010) and to use the best predictor to forecast the discharge of the river from 2010 to 2020. The performance of the two models was subjected to statistical test based on correlation coefficient (r) and the root‐mean‐square error. The result showed that autoregressive integrated moving average performs better considering the level of root‐mean‐square error and higher correlation coefficient.
Effect of attendance on performance in postgraduate courses in science and engineering
(International Centre for Mathematical & Computer Sciences, Lagos, Nigeria., 2012) Udomboso, C. G.; Falode, A. O.; Chukwu, A. U.
Most postgraduate students in developing countries like Nigeria are working class students that have to shuttle between their workplaces and classes. Reason being that there few or no sponsors for postgraduate programmes exist in the country. Therefore, most students are self sponsored. Furthermore, most postgraduate courses in Nigeria are full time programmes. There are some core courses that require strict attendance in classes. However, it is seen that this is not always the case as the students have to attend to their jobs as well; otherwise they might lose the jobs and have no fund to continue. The programmes considered in this study are those in Statistics and Petroleum Engineering. This study therefore looks at the effect of students’ attendance in postgraduate classes to their performances, and also proffers solutions to its long-term effects on the industrial and economic developments.
Effect of attendance on performance in postgraduate courses in science and engineering
(International Centre for Mathematical & Computer Sciences, Lagos, Nigeria., 2012) Udomboso, C. G.; Falode, A. O.; Chukwu, A. U.
Most postgraduate students in developing countries like Nigeria are working class students that have to shuttle between their workplaces and classes. Reason being that there few or no sponsors for postgraduate programmes exist in the country. Therefore, most students are self sponsored. Furthermore, most postgraduate courses in Nigeria are full time programmes. There are some core courses that require strict attendance in classes. However, it is seen that this is not always the case as the students have to attend to their jobs as well; otherwise they might lose the jobs and have no fund to continue. The programmes considered in this study are those in Statistics and Petroleum Engineering. This study therefore looks at the effect of students’ attendance in postgraduate classes to their performances, and also proffers solutions to its long-term effects on the industrial and economic developments.
Factors affecting learning in an open and distance learning programme
(2010) Dontwi, I. K.; Amahia, G. N.; Chukwu, A. U.; Udomboso, C. G.
There is bound to be a shift towards those courses that will provide the knowledge and skills for economic relevance and earning power. Commerce, science and technology are likely to be oversubscribed, once driven world, seems to be diminishing steadily. When designing instruction for distance education, attention is often focused on the cognitive domain, as it is in "traditional" (face-to-face) instruction. What do the students need to know? Which instructional strategies will be most appropriate? Upon what performance criteria will learners be evaluated?
Introduction to probability and stochastic processes with applications
(Fasco Publishers, 2014) Shittu, O. I.; Otekunrin, O. O.; Udomboso, C. G.; Adepoju, K. A.
Modeling students’ academic performance using artificial neural network
(Federal University, Ndufu-Alike Ikwo (FUNAI), Nigeria, 2016) Asogwa, O. C.; Udomboso, C. G.
Artificial Neural Network has been discovered as a better alternative to traditional models and that is why a model based on the Multilayer Perceptron algorithm was developed in this study. The appropriate number of hidden neurons that best modeled the academic performance of students was determined by the developed Network algorithm. Test data evaluation showed that Network Architecture 17-80 -1 was chosen among the numerous developed network architectures because of its model performances. The chosen network architecture gave the minimum value of Mean Square Error (MSE = 0.0718), minimum value of Network Information Criteria (NIC = 0.0743), maximum value of R- Square (R2=0.8975) and maximum value of Adjusted Network Information Criteria (ANIC= 0.8931). It was equally observed that there were patterns in the movement of hidden neurons against the model evaluation criteria. As the number of the hidden neurons appreciates the value of both MSE and NIC decreases down the plot, while that of ^-Square and ^MCvalues appreciate down the plot. The network was able to model the research problem with acceptable values judging from the model checking criteria considered in this work. Also the order of contribution of the predictor variables to the model was determined.
Modelling trends in contraception usage in Nigeria and Ghana
(SAGE Publications, 2018) Udomboso, C. G.; Amoateng, A. Y.
This study examines trends in contraception usage using Demographic and Health Survey (DHS) data from Nigeria (2013) and Ghana (2008, 2014). We used a cubic spline to estimate values between intervals, analysed using the time-series neural network model and forecasting till 2030. Results show contraception usage increasing with an average rate of 4.4 per cent, desire not to use declining at an average rate of 0.7 per cent and the use of modern contraception increasing at an average rate of 5.1 per cent. Use of traditional contraception is still increasing in Nigeria at a rate of 0.6 per cent but declining in Ghana at a rate of 0.3 per cent. By 2030, injectables would be mostly used, while the withdrawal method will be preferred among those still in favour of the traditional method. These trends show the readiness of the two countries to embrace the use of modern contraception in an effort to promote the campaign for Planned Parenthood and Family Programme.
Neural network regression for modelling the effects of selected soil physico-chemical properties on adsorption
(Nigeria Statistical Society, 2017) Udomboso, C. G.; Nzelu, N.; Olu-Owolabi, B. I.
Heavy metals in soils have been known as soil pollutants, to constitute serious economic importance as their accumulation has led to reduced agricultural production and quality of life. In the present paper, we studied the adsorption behaviour of selected heavy metals in soils, due to some physico-chemical properties. The soil under study was obtained from the River Benue Basin in the middle belt region of Nigeria. The heavy metals considered included lead (Pb), zinc (Zn), copper (Cu), and cadmium (Cd), while the physico-chemical properties included hydrogen ion concentration (pH), percentage goethite, percentage humic acid, time, and sorbate concentration. Estimation of the effects was carried out using the statistical neural network at α = 0.05, while the cubic spline was used to interpolate within values and extrapolate forecasted values. Results show that rates of adsorption differ across properties. In all physical properties, except humic acid, Cd is most adsorped at AIC of 0.067, 0.079, 0.002, and 21.137 (all at p<0.05). For humic acid, most adsorped is Zn at AIC of 5.692 (p<0.05). These call for effective soil management system in Nigeria, which is expected to yield reliable data on soil behaviour, as well as concerted effort in eradicating (or reducing) the presence of these pollutants.
On R2 contribution and statistical inference of the change in the hidden and input units of the statistical neural networks
(Society of African Journal Editors, 2012-11) Udomboso, C. G.; James, T. O.; Odim, M. O.
Determining the number of liitltlen units for obtaining optimal network performance has been a concern over the years ilespite empirical results showing that with higher neurons, the netivork error is retlucetl. This has led to indiscrimate increase in the hidden neurons, thereby bringing about overfitting. On the other hand, using too few hidden neurons leads to error bias, which can make neural network statistically unfit. In this paper, we developed a model for R1 for investigating changes in hidden and input units, as well as developed tests that can be used in determining the number of hidden and input units to obtain optimal performance. The result of the analyses shows that there is effect on the network model when there is an increase in the number of hidden neurons, as well as the number of input units.
On some properties of a hetereogeneous transfer function involving symmetric saturated linear (SATLINS) with hyperbolic tangent (TANH) transfer functions
(JMASM, Inc., 2013-11) Udomboso, C. G.
For transfer functions to map the input layer of the statistical neural network model to the output layer perfectly, they must lie within bounds that characterize probability distributions. The heterogeneous transfer function, SATLINS_TANH, is established as a Probability Distribution Function (p.d.f), and its mean and variance are shown.
On the level of precision of the wavelet neural network in rainfall analysis
(2014) Udomboso, C. G.; Amahia, G. N.; Dontwi, I. K.
This research combines the efficiency of the artificial neural network and wavelet transform in modelling rainfall. The data used were decomposed into continuous wavelet signals on a scale of 48. Each of the decomposed series was subjected to correlation test with the original data. Instead of using all the series, ten series were selected on the basis of high correlation with die original data. These series included CWT 1, CWT 2, CWT 4, CWT 3, CWT 6, CWT 8, CWT 5, CWT 10, CWT 12, and CWT 7 (according to rank). The analysis showed that except in extremely rare cases, all the series performed optimally compared to the original data. The result of the study has been able to show' that using the continuous w'avelet transform in the ANN technique, a better performance of the network is observed.
On the maximization of the likelihood function against Iogarithmic transformation
(2008) Obisesan, K. O.; Udomboso, C. G.; Osowole, O. I.; Alaba, O. O.
We consider maximum likelihood estimation logarithmic transformation irrespective of mass of density functions. The estimators are assumed to be consistent, convergent and existing. They are referred to as asymptotically minimum-variance sufficient unbiased estimators (AMVSU). We find that the likelihood function gives accurate result when maximized than the log-likelihood. This is because logarithmic transformation has potential problems. We consider a uniform case where the parameter 0 cannot be estimated by calculus but order-statistics. We fit a truncated Poison distribution into data on damaged done after estimating λ by a Newton-Raphson Iterative Algorithm.