Iraqi Statisticians Journal

Chaotic time series forecasting such as maximum wind speed rates is of great importance in the fields of meteorology and renewable energy to reduce and control the harmful negative effects. The problem of wind speed is that it is affected by several interrelated factors such as temperature and atmospheric pressure, which are characterized by non-linearity through the influence of time series on differences that may be a cause of the emergence of uncertainty problems, which makes it difficult to model using traditional univariate time series methods. Echo State Network (ESN) is a neural network specialized in time series forecasting after addressing the problem of synchronization with the time variable as a recurrent network to address time-dependent effects and accurate prediction of time series in addition to its ability to model nonlinearly. This study presents the use of the Autoregressive (AR) model and then its hybridization with the deep echo state network, which is called the AR-ESN hybrid method by using the optimal structure of the AR model to determine the optimal inputs to the ESN network as the main contributions to solving the prediction problems for real data forecasts. The use of ESN as a proposed forecasting method is to improve the forecasting efficiency to reduce the risks associated with extreme weather fluctuations compared with traditional forecasting results. The results indicate that the ESN model based on AR model can contribute to increasing the forecasting accuracy of maximum wind speed compared with traditional models by using mean absolute percentage error (MAPE) as one of the criteria the forecasting accuracy.

Studying climate in terms of predicting evaporation quantities through changing many factors and climate variables is of great importance to reduce the risks of climate change and its impact on many environmental phenomena affecting human and plant health. Weather patterns and conditions will be analyzed for a limited period of time. In this study, Ridge Regression (RR) was used to identify components, address the problem of multicollinearity, and overcome the problem of multicollinearity of the reconstructed inputs. The problem of multicollinearity occurs when building a multiple linear regression model, i.e. the presence of a strong correlation between the predictive variables, and this correlation leads to an increase in the variance of the regression parameter estimates and obtaining unstable estimates. Forecasting chaotic time series represented by climate data affecting the evaporation time series often suffers from the problem of nonlinearity in addition to multicollinearity. The Echo State Network (ESN) is a neural network specialized in predicting time series after addressing the problem of synchronization with the time variable as a recurrent network to address time-dependent effects and accurate prediction of time series in addition to nonlinear modeling. RR provides accurate predictions due to its structural effect to avoid overfitting while ESN will address the nonlinearity problem in addition to the problems addressed by RR model which ESN structure is based on. The expected results show that ESN model based on RR significantly outperforms traditional models in multivariate time series forecasting.

During the past few years, the number of classification techniques has increased with the rapid growth of technology, which depends on machine learning. Recently, medical experts and doctors have widely utilized machine learning, a branch of artificial intelligence, to aid in predicting and diagnosing various diseases. Regarding health research, machine learning techniques are used extensively for data processing. In this study, we applied three different machine learning algorithms to a medical diagnosis problem and analyzed their efficiency in predicting the results. The study focuses on the diagnosis and factors influencing renal failure disease, using a serum test for both presence and absence patients. The dataset used for the study consists of 165 cases and 13 attributes of RFD patients. The goal of this study is to find out how well k-nearest neighbors (KNN), decision tree (DT), and random forest (RF) classifiers work by looking at things like accuracy, geometric mean, kappa coefficient and area under the curve for RFD prediction. The experimental results show that the RF classifier performs better than the other classifiers. Additionally, based on the final fitted models, it was found that urea, albumin, and magnesium are the most significant factors that clearly impact patients with renal failure disease.

Satellite imagery often suffers from noise due to various atmospheric and sensor-related factors, which can significantly degrade image quality and hinder subsequent analysis. This article presents a comprehensive study on denoising satellite images utilizing spline interpolation and exponential spline techniques. We propose a novel nonlinear filter designed to enhance the denoising process, and we compare its performance against traditional spline-based methods. The evaluation metrics employed include the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM), which are critical for assessing image quality. Our experimental results demonstrate that the proposed nonlinear filter consistently outperforms both spline interpolation and exponential spline methods, achieving superior PSNR and SSIM values. This study highlights the proposed filter's effectiveness in preserving image details while reducing noise and contributes to the ongoing advancements in remote sensing image processing techniques. The findings underscore the potential of nonlinear filtering approaches in enhancing the quality of satellite imagery for various applications.

Image denoising is one of the fundamental aspects of removing noise from an image and enhancing its features containing visual information. Based on this, Convolutional Neural Networks (CNNs) have been a latest topic of study, with a wide range of applications in fields as diverse as diagnostic image denoising and low-light image denoising. In this paper, an image denoising method is proposed based on converting the noisy image to YUV colon space, extracting the noisy Y channel, and obtaining an appropriate smoothing parameter for the noisy Y channel using the cross-validation smoothing techniques that are used to estimate the Y density function, then using an appropriate noise reduction method (total variation denoising) on the estimated density function to extract the denoised Y channel by removing the Gaussian noise. The results showed that the new proposed method effectively removed noise from the image, which is attributed to the approach adopted in this proposed filter.

Survival models play a key role in analyzing time-to-event data. Accelerated time models are useful when the effect of the covariates on the survival time is proportional throughout the follow-up period. The well-known Weibull accelerated failure-time model (AFT), as an accelerated time model, is widely employed in survival analyses. The high-dimensional variables picked, considering most of the low-dimensional AFT model-based variable selectors, assume that the covariate effects are constant throughout the period. However, the Weibull AFT model includes time transformations enabling constant, increasing, or decreasing impacts during the entire study, and thus, it is a more flexible approach. Lemurs’ optimization algorithm (LOA), a new powerful algorithm, is employed in the covariate selection. It is important to select appropriate significant covariates for AFT models in practice. However, like all other high-dimensional data, the lemur faces the curse of dimensionality problem in Weibull AFT model-based variable selection. Thus, it is necessary to have a purposeful variable selection algorithm for Weibull AFT models that considers the increasing, constant, or decreasing effects of covariates.