IRAQI JOURNAL OF STATISTICAL SCIENCES

Comparison of Wavelet Shrinkage and Hampel Filter in the Analysis of Multivariate Linear Regression Models

2025-11-26T07:02:38+00:00

The presence of outliers in the data of a multivariate regression model affects the accuracy of the estimated model parameters and leads to unacceptably large residual values. Therefore, some filters, including the Hempel filter, are usually used to handle outliers (or use some robust method). This paper proposes to employ wavelet shrinkage to address the problem of outliers in multivariate regression model data by using wavelets (Coiflets, Daubechies, and Demy) with a universal threshold method and soft rule. To illustrate the efficiency of the proposed method (Wavelet Shrinkage filter) was compared with the traditional method (Hampel filter) based on the mean square error criterion through simulation and real data. A program has been designed in MATLAB to do this. The results proved that the Wavelet shrinkage filter method was more efficient than the traditional method in dealing with the outlier problem and obtaining more accurate multivariate model parameters than the Hampel filter method.

Using Deep learning and GIS applications to Extract Features from Remote Sensing Data

2025-11-26T07:09:12+00:00

In recent years, artificial intelligence (AI) has advanced quickly, equal or perhaps even outperforming human accuracy in tasks like picture recognition, reading comprehension, and text translation. Large-scale opportunities that were not previously available are now had been created by the confluence of AI and GIS with remote sensing data processing. The motivation of current study is to incorporate deep learning models that implemented through ArcGIS Pro tools particularly convolutional neural networks (CNNs), identifying complex patterns and features in high-resolution image. An automated Deep Learning model type Mask R-CNN had applied to extract model for training objects. The overall accuracy metric improved the performance of the current work accurately with less error when calculating RMSE criteria.

Diagnosis and Classification of Alzheimer's Disease Using Some Machine Learning Models: A Comparative Study

2025-11-26T07:14:11+00:00

Alzheimer's disease damages brain neurons, resulting in memory loss. Early and accurate diagnosis of the disease is crucial for implementing preventive measures. However, differentiating between Alzheimer’s and healthy data in older adults is challenging due to the similarities in their brain patterns and intensities, complicating researchers' efforts to make an accurate diagnosis. Therefore, the research aims to use machine learning to improve diagnosis and classification of the disease, such as support vector machines (SVM), decision trees, and feedforward neural networks (FFNN). Classification algorithms were applied to the Alzheimer’s disease dataset, including 2149 cases, and the models were evaluated through metrics (Accuracy, Precision, Recall, specificity, F1 Score, F2 Score, F3 score, and AUC).

Following data analysis and obtaining the results, we reached the Decision Tree model excels across all metrics, achieving high scores in accuracy (96.32%), precision (94.63%), recall (95%), specificity (97.05%), and AUC (94.96%). This demonstrates its ability to correctly identify true positives and negatives, and reduce false positives and negatives, makes it the most reliable model for accurately classifying Alzheimer's disease cases. In contrast, the SVM linear and FFANN models offer a good balance with accuracy (83.53% and 83.57%), specificity (89.27% and 91.72%), and AUC (89.63% and 89.84%). However, their lower recall (73.03% and 68.68%) compared to the Decision Tree may result in missed positive cases, making them less effective for classification. The SVM RBF model is the least effective option, with high precision and specificity but poor performance across all other metrics and lacks overall balance, resulting in a high number of false negatives. metrics and lacks overall balance, resulting in a high number of false negatives.

Conclusion: The decision tree model outperforms other models, making it the best choice for Alzheimer's disease classification.

Estimating the Population Mean in Stratified Median ranked Set Sampling Using Combined and Separate Regression with the Presence of Outliers

2025-11-26T07:32:21+00:00

This research aims to demonstrate the high efficiency and accuracy in estimating the limited population mean through the estimates of the separate and combined stratified regression line based on the method of median ranked set sampling to choose a sample that is more representative of the community. With the problem of heterogeneity in the data and containing extreme values (outliers), it is recommended to use stratification of the community and draw samples using the method of sampling the middle ordered from these layers, which is known as the stratified median ranked set sampling ( ), which is one of the modified ranked set sampling methods (RSS), where the mean square error (MSE) of the population mean estimator obtained in this way was compared with the MSE value of the population mean estimator obtained through regression estimates using the robust variance and covariance matrix (Minimum Covariance Determinant (MCD), Minimum Volume Ellipsoid (MVE)) to calculate the averages and using robust methods (Huber M, Huber MM, Least Median of Squares (LMS), Least Trimmed Squares (LTS)) to estimate the regression parameter. The simulation results show that the proposed estimator outperforms the robust estimators in most cases because it obtains the lowest values of the mean square error.

A Multi-Objective Optimization Approach for Design Earth-Fill Dams Variables

2025-11-26T07:46:30+00:00

The entire research study is about the optimization of design variables for an earth-fill dam utilizing multi-objective optimization approach. We have considered three variables that are related to the soil material properties which are core density, foundation Young's modulus of elasticity, and shell coefficient of permeability. A limited range of each variable is applied depending on the literature and numerical simulations are dedicated for the response of the global structure using ABAQUS program. Box-Behnken design method along with MATLAB codes with least-square method are being used to construct two surrogate models for the displacements of the earth-fill dam. Fifteen numerical models are involved in the process with the presence of non-linear equations for the objective function for the optimization process. The objective functions were for the pore water pressure and the maximum principal stress were they would be first checked for reliability by the coefficient of determination check R². The reliability of the objective function was 100% which enhanced them to be ready for the multi-objective optimization step. The results of the optimized variables for both objective functions were determined and compared with the minimal responses of the considered models of the numerical simulations of the global structure. The optimum results of both objective functions of the earth-fill dam were determined and approved which is an indication of excellent result for the optimization of the design variables.

Diagnosing Cross-Sectional Time Series Models of Money Supply For Arab Countries

2025-11-26T08:05:51+00:00

Using cross-sectional time series models, the study concludes by determining the extent of development of trends in fiscal policy tools and money supply in the narrow sense for three countries (Iraq, Jordan, and Algeria). It also suggests a statistical model to analyze the study data and determines the extent of the impact and effectiveness of fiscal policy tools on money supply through its financial tools represented by public spending and tax revenues. Eviews10, a statistical tool, was used to handle annual data for the years 1993–2023. The cumulative regression model is the best model, according to the results of the Fisher statistical test. The money supply in the countries under investigation has a substantial positive relationship with tax revenues, whereas public spending has a non-significant positive relationship. The estimated parameters of the suggested model are consistent with both practical reality and economic theory presumptions.

Ranking of Intuitionistic Fuzzy Numbers by Using Scaling Method

2025-11-26T08:10:24+00:00

Ranking intuitionistic fuzzy numbers (IFN) is a challenging task. Several methods have been presented for ranking IFNs. Also ranking for three IFN is rare. In this work, a new multidimensional scaling (MDS) method for ranking triangular intuitionistic fuzzy number (TIFN) is proposed. This method is easy to implement, visualized and embedded the (TIFN). Also, gives a possibility to configure points in different ways. Configuration points can be extracted in a two-dimensional space since each TIFN is represented as a row in a matrix. Since these points are not uniquely established, we provide a technique for reconfiguring it in order to compare it with various methods. This method is novel in sense of the idea. Lastly, the method is illustrated through numerical examples.

Integrating Wavelet Shrinkage with SURE and Minimax Thresholding to Enhance Maximum Likelihood Estimation for Gamma-Distributed Data

2025-11-26T08:15:36+00:00

This paper uses the Maximum Likelihood Estimation method to investigate the impact of data contamination on the accuracy of parameter estimation for the Gamma distribution. A de-noising approach based on wavelet shrinkage has been proposed to address the limitations posed by contamination. Several types of wavelet functions were employed in combination with different threshold estimation techniques, namely Universal, Minimax, and Stein’s Unbiased Risk Estimate, applying the soft thresholding rule. The study involved simulating data sets generated from the Gamma distribution and analyzing real-life data assumed to follow the same distribution. A specialized program was developed in MATLAB to conduct these simulations and implement both the classical Maximum Likelihood Estimation method and the proposed wavelet-based de-noising techniques. The performance of the parameter estimates was compared using the Mean Squared Error criterion. The findings demonstrated that data contamination significantly affects the accuracy of parameter estimates obtained through the classical Maximum Likelihood Estimation method. In contrast, the proposed wavelet shrinkage method effectively reduced the influence of contamination and enhanced the accuracy of parameter estimation for the Gamma distribution. The study highlights the practical value of integrating wavelet-based denoising techniques into statistical estimation processes, particularly when working with contaminated datasets.

A Proposed Method Based on Logistic Regression and Cluster Analysis in Selecting Influential Variables for Kidney Failure Patients

2025-11-26T08:21:34+00:00

The research aims to study kidney failure by analyzing the relationship between it and a set of independent variables. To achieve this, a method was proposed that relies on reducing the number of independent variables used in binary logistic regression. The method relies on merging the independent variables with the dependent variable using cluster analysis, to improve the accuracy of the model and obtain the best possible results.The proposed method was applied to a sample of 142 individuals to study the relationship between the response variable (renal and non-renal failure) and independent variables such as gender, age, smoking, urea, creatine, and calcium. The results showed that the proposed method succeeded in reducing the number of independent variables and provided an ideal model that classifies the data with high accuracy. The resulting model focused on the two most influential variables, urea and creatine, and achieved a high classification rate of 94.4%.,The proposed method proved effective in reducing the number of variables and achieving accurate results in classifying data related to kidney failure.

A New Theorem for Lower Bounds in NP-Hard Multi-Objective Scheduling Problems

2025-11-26T08:27:23+00:00

On a single machine, each of n jobs must be processed continuously. At time zero, every job is ready for processing. The tasks to process a sequence that minimizes the total sum of competition times plus the sum of tardiness . This bi-criteria problem is NP-hard because of the second one. We provide a theorem that demonstrates a relationship between the optimal solution, lower bounds, and the number of efficient solutions. The case is that the theorem works for NP-hard problems, whereas in previous works the focus was on P-hard problems. The theorem limits the lower bound's range, which is crucial for determining the best answer. Additionally, the theorem allows for discovering new lower bounds by opening algebraic procedures and concepts.

Fuzzy Discriminant Analysis with Application

2025-11-26T08:35:05+00:00

In the current research, two methods of data classification were used, namely; the Linear Discriminant analysis (LDA) and the Fuzzy Discriminant Analysis (FDA). The discriminant analysis is considered as a method in which a certain item is analyzed into one of the sets depending on statistically significant variables in the process of analysis. The item is classified within the suitable population through a set of variables. Sometimes, the data is non-linear or does not follow a normal distribution or is unstable, so the use of fuzzy logic is successful because it does not require the data to follow a normal distribution or to be stable and Through fuzzy theory, models are developed to solve problems that cannot be analyzed using pure mathematical methods. The research aims to compare the two analysis methods (LDA) and (FDA) in order to display data and distinguish between different categories with high accuracy, provide accurate classifications of observations, as well as diagnose variables of high importance in discriminatory data using confidence intervals for the Roy-Bose test and the t-test. Real data was used for two groups of kidney patients (infected, not infected) depending on a set of influencing factors and the extent to which each factor affects the persons or the new items and their distribution on one of the two sets, which in turn, leads to early diagnosis and avoiding the deterioration of the health state of the patient. Results were obtained after applying the statistical package (SPSS) and (R) statistical package, which indicated that the fuzzy discriminant analysis (FDA) was the best analysis as it yields the least classification error by depending on the mean standard error (MSE).

Finding the Optimal Prediction of the Occurrence of Earthquakes Using Markov Chains and Artificial Intelligence Methods

2025-11-26T08:45:28+00:00

In this research, a method for predicting earthquakes was presented, where seismic data for the Syrian coastal region were studied for the period from 1996 to 2011, which included the earthquake strength, intensity, location and date of occurrence. In light of the analysis of this data, a method for prediction using Markov chains was determined for a specific period. After that, this process was improved using one of the artificial intelligence methods, which is the machine learning method using the random forest method. Due to this improvement, better accuracy in the results was obtained.The research presents an advanced approach to earthquake predictions by integrating artificial intelligence techniques with Markov chains, thus obtaining the proposed predictions that enable communities to take better preventive measures and remove risks from the population. This will lead to reducing losses in lives, property and infrastructure as much as possible. This is the goal of all studies related to predicting earthquakes in the world and of all types

Application of Advanced Statistical Models in Big Data Analysis: Modern Methodologies and Techniques

2025-11-26T08:55:21+00:00

This examine investigates the software of advanced statistical fashions in huge records analytics, that specialize in their capability to cope with demanding situations in accuracy, scalability, and moral alignment inside records-driven choice-making.
The research employs a multi-faceted method, integrating Gradient Boosting Machines (GBM) with hyperparameter tuning for credit chance prediction, Bayesian Reinforcement Learning (BRL) for dynamic uncertainty modeling, and quantum computing simulations for optimization responsibilities. Distributed computing frameworks, including Kubernetes, and privacy-retaining techniques like homomorphic encryption are evaluated to decorate computational and ethical robustness. The GBM version carried out a 20% discount in category blunders in comparison to conventional methods, whilst BRL proven superior interpretability in stochastic environments. Real-time adaptive fashions reduced latency through 60% in streaming facts situations, and quantum-greater algorithms showed a 75% development in dimensionality reduction performance. Ethical frameworks, such as adverse debiasing, decreased demographic parity gaps from 15% to three% without compromising model overall performance. The findings recommend for hybrid models that merge statistical intensity with computational innovation, emphasizing their essential position in overcoming scalability and bias challenges. Future studies must prioritize quantum-geared up architectures and interdisciplinary methodologies to maintain improvements in big records analytics. This work contributes a foundational framework for deploying statistically rigorous, ethically aligned, and computationally efficient solutions in complex facts ecosystems.

Bayesian Estimation of the Inverse Rayleigh Process under a Non-Homogeneous Poisson Process Framework

2025-12-01T07:33:32+00:00

The rationale on which this study is based is that accurate and dependable means to obtain time-dependent failure rates in repairable systems, especially in cases that are not homogeneous, are required, and the conventional models are not always in a position to meet these demands. To address this, the research targets at use of Inverse Rayleigh Process (IRP) within a Non-Homogeneous Poisson Process (NHPP) paradigm, a model of the system failures that suits the use of the stochastic model of a system. To improve the accuracy of parameter estimation, the Maximum Likelihood Estimation (MLE) approximation and Bayesian methods are studied and here the solving of the analytical problems due to intractable posterior distributions when using Laplace approximations is sought. Zooming over the simulation experiments that have been conducted on various sample sizes, evaluated through Root Mean Square Error (RMSE), shows that the Bayesian estimator in particular Bayes II prior outperforms MLE. Lastly, the proposed approaches are confirmed on the real-life failure records in the Mosul Gas Power Plant, which confirms the effectiveness of the Bayesian approach in the modeling of the coupled reliability systems in practice and more precisely in the data-scarce context

Shrinkage Estimators in Bell Regression Model: Subject Review

2025-11-26T08:31:57+00:00

Bell regression model has become a very versatile model that replaced the conventional count data models and helps to resolve the problem of over-dispersion where the variance of data points surpasses the mean. Nevertheless, in practice, the classical maximum likelihood estimators (MLE) of the parameters of a model are frequently affected by multicollinearity among the explanatory variables, and they yield highly unstable estimates and inflated variances. To address these difficulties, estimation methods developed to estimate shrinkage, such as ridge or Liu estimators have been applied to the Bell regression model. In the subject review, new developments in estimators of shrinkage of Bell regression models are proportionate in discussing their theoretical background in knowledge, estimation process, and asymptotic characteristics. The results on Monte Carlo simulation studies always show that shrinkage estimators outweigh MLEs in that they minimize mean squared error and bias more than MLEs, especially in cases of multicollinearity. Both overall, estimation methods of shrinkage is a considerable improvement in Bell regression modeling that offers certainty and effectiveness of analysis of complicated counts information utilizing problematic distribution attributes.

Skewness Measures – Article Review

2025-11-26T08:48:51+00:00

The research aims to define skewness to know the type of skewness, by applying skewness measures such as the Karl-Pearson measure in its cases (median, median and moments) and Kelly’s measure and Bowly’s measure on real data for the variable (x, y), where the variable (x) represents the data for cement expansion taken from Badoush Cement Plant n=80 for the period (2008-2010) (Al-Sarraf, 2021) and the variable (y) represents data showing the times of failure resulting from various reasons for the life test device for n=30 devices (ZEENALABIDEN, 2023) .The Z-criterion was used to measure skewness for the data of variable X and for the data of variable Y. The results for both variables were outside the acceptable period for this criterion (1.96, -1.96).