Journal of Intelligent Data Analysis and Computational Statistics (p-ISSN: 3049-3056 e-ISSN: 3048-7080)

Correlation Analysis in Multidisciplinary Research: A Systematic Review of Theoretical Foundations, Methodological Frameworks, and Empirical Applications

Fri, 01 May 2026 00:00:00 +0000

Correlation analysis is a method researchers use to determine whether two variables are related. In this review, eight studies from various fields, including education, healthcare, finance, and social sciences, were examined. The goal was to understand how correlation is used and what kind of results it produces. Most researchers used Pearson’s correlation to measure relationships. They also used tests such as the t-test and Fisher’s z, often with SPSS software, to see if the results were statistically meaningful. In one education study, there was a moderate positive relationship (r = +0.34) between students’ awareness of nature and their science performance. This suggests there may be some link. However, studies related to online learning did not always show a significant connection. It is important to remember that correlation does not prove cause and effect. Two variables may be related without one directly causing the other. Overall, correlation is useful, but it must be interpreted carefully.

AI-based Cattle Disease Detection System Using Modified CNN Architecture for Blood Smear Image Analysis

Sravani P, Mariya Sneha T, Shiva Sumanth Reddy — Tue, 05 May 2026 00:00:00 +0000

Tick-borne diseases such as babesiosis and anaplasmosis pose significant threats to cattle health, causing substantial economic losses in livestock farming. Traditional microscopic diagnosis of these diseases through blood smear analysis is time-consuming, labor-intensive, and prone to human error. This study presents an AI-based cattle disease detection system utilizing a modified convolutional neural network (CNN) architecture for automated blood smear image analysis. The proposed system implements a three-tier architecture comprising farmer, lab technician, and veterinary doctor modules, enabling seamless coordination in disease diagnosis. This modified CNN architecture incorporates optimized convolutional layers with ReLU activation, max-pooling strategies, and dropout regularization to enhance feature extraction from microscopic blood smear images. The system achieved 98.0% accuracy for babesiosis detection and 97.7% accuracy for anaplasmosis detection, with precision and recall exceeding 95%. Experimental results demonstrate that the proposed approach significantly outperforms traditional diagnostic methods and baseline CNN models, providing rapid, accurate, and cost-effective disease detection for improved cattle health management.

DT-SVM and Hybrid Approaches for Missing Data Imputation and Classification: A Comprehensive Survey

Satish Kumar Kalagotla, Thoudam Basanta, Mutum Bidyarani Devi — Fri, 01 May 2026 00:00:00 +0000

Missing data represents a pervasive challenge in real-world datasets, particularly within medical research and clinical applications, where its presence can substantially degrade the performance of machine learning classifiers and compromise the validity of analytical conclusions. This comprehensive survey paper systematically examines hybrid approaches that integrate decision trees (DT) and support vector machines (SVM) for missing value imputation and subsequent classification, with particular emphasis on the DT-SVM framework and its algorithmic variants. The study provides a thorough exploration of missing data mechanisms, evaluates traditional and machine learning-based imputation techniques, and delineates the theoretical foundations of decision trees and support vector machines. Through critical analysis of existing hybrid methodologies and comparative evaluation against conventional approaches, this review synthesizes current literature to reveal that DT-based imputation, which leverages enhanced attribute correlations within homogeneous data segments identified through recursive partitioning, consistently outperforms simple imputation methods when combined with SVM classification. The survey further examines recent advancements, including approximated k-nearest neighbor (A-kNN) variants that address computational efficiency concerns while maintaining classification accuracy. Key research gaps are identified, including challenges in high-dimensional settings, handling of missing not at random mechanisms, and integration with deep learning architectures. The findings collectively suggest that integrated frameworks such as DT-SVM represent a promising trajectory for achieving robust classification performance in the presence of missing data, with particular relevance to medical diagnosis applications where data quality issues are prevalent and prediction accuracy is paramount.