Gurram Sravan Kumar, Dr. Arathi Chitla

DOI: https://doi.org/10.5281/zenodo.24128940

This study presents a reproducible, source-aware data audit and exploratory data analysis framework for a hybrid Coronary Artery Disease (CAD) dataset constructed from four sources: Cleveland, Hungarian, Switzerland, and Long Beach VA. The study focuses on dataset provenance, schema validation, target coding, missing-value analysis, repeated clinical records, invalid-value checks, feature distributions, categorical frequencies, relationships between clinical attributes, and exploratory imputation strategies. The dataset comprises 920 records with 13 clinical predictors and a binary target, exhibiting substantial missingness in key attributes and considerable variation in target prevalence across sources. By systematically examining overall and source-specific data characteristics, this study establishes a transparent data-quality and preprocessing foundation for subsequent machine learning research. The findings highlight the importance of preserving source information and addressing data-quality challenges before predictive modelling, without making claims about predictive performance or clinical readiness.

Coronary Artery Disease, UCI Heart Disease Dataset, Source-Aware Data Audit, Hybrid Dataset, Missing Data, Exploratory Data Analysis, Data Provenance, Data Quality Assessment.

This paper presents a source-aware data-audit and exploratory data analysis (EDA) framework for a 920-record hybrid Coronary Artery Disease (CAD) dataset assembled from the Cleveland, Hungary, Switzerland, and Long Beach VA sources. The dataset contains 13 clinical predictors, a binary target, the original source-specific target label, and source metadata. The audit identified substantial missingness in ca, thal, and slope, an overall target distribution of 411 class-0 and 509 class-1 records, and considerable variation in target distributions across the four sources. Two repeated clinical signatures were identified by comparing the 13 clinical predictors and the binary target. Following the documented duplicate-handling procedure, two records were removed, leaving 918 records for subsequent analysis. The plausible-range checks identified no observations outside the specified screening limits, while the interquartile-range procedure flagged some observations for review without automatically removing them. The exploratory imputation comparison demonstrated that both mean-plus-most-frequent and median-plus-most-frequent strategies could produce predictor matrices with no remaining missing values. However, this comparison did not establish that either strategy was superior for predictive modelling. The findings highlight the importance of retaining source information, documenting data-quality decisions, and evaluating missing-value handling before developing machine-learning models. Future work will focus on predictive modelling using suitable validation procedures, including Leave-One-Source-Out evaluation, to examine generalization to a source excluded from model training. All preprocessing steps, including imputation and feature selection, should be performed within the training process to avoid information leakage. This study establishes a transparent data-quality and EDA foundation for subsequent CAD prediction research. It does not establish predictive performance, generalization to unseen sources, or clinical utility.

[1] A. Janosi, W. Steinbrunn, M. Pfisterer, and R. Detrano, “Heart Disease,” UCI Machine Learning Repository, 1989. DOI: 10.24432/C52P4X. https://archive.ics.uci.edu/dataset/45/heart+disease
[2] R. Detrano et al., “International application of a new probability algorithm for the diagnosis of coronary artery disease,” The American Journal of Cardiology, vol. 64, no. 5, pp. 304–310, 1989. DOI: 10.1016/0002-9149(89)90524-9. https://pubmed.ncbi.nlm.nih.gov/2756873/
[3] J. A. C. Sterne et al., “Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls,” BMJ, vol. 338, b2393, 2009. DOI: 10.1136/bmj.b2393. https://pmc.ncbi.nlm.nih.gov/articles/PMC2714692/
[4] M. K. Hasan et al., “Missing value imputation affects the performance of machine learning: A review and analysis of the literature (2010–2021),” Informatics in Medicine Unlocked, vol. 27, 100799, 2021. DOI: 10.1016/j.imu.2021.100799. https://doi.org/10.1016/j.imu.2021.100799
[5] C. L. Andaur Navarro et al., “Risk of bias in studies on prediction models developed using supervised machine learning techniques: systematic review,” BMJ, vol. 375, n2281, 2021. DOI: 10.1136/bmj.n2281. https://www.bmj.com/content/375/bmj.n2281
[6] R. Alizadehsani et al., “A data mining approach for diagnosis of coronary artery disease,” Computer Methods and Programs in Biomedicine, vol. 111, no. 1, pp. 52–61, 2013. DOI: 10.1016/j.cmpb.2013.03.004. https://pubmed.ncbi.nlm.nih.gov/23537611/
[7] G. S. Collins et al., “Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement,” BMJ, vol. 350, g7594, 2015. DOI: 10.1136/bmj.g7594. https://www.bmj.com/content/350/bmj.g7594
[8] S. W. J. Nijman et al., “Missing data is poorly handled and reported in prediction model studies using machine learning: a literature review,” Journal of Clinical Epidemiology, 2021. DOI: 10.1016/j.jclinepi.2021.11.023. https://pubmed.ncbi.nlm.nih.gov/34798287/
[9] S. van Buuren and K. Groothuis-Oudshoorn, “mice: Multivariate Imputation by Chained Equations in R,” Journal of Statistical Software, vol. 45, no. 3, pp. 1–67, 2011. DOI: 10.18637/jss.v045.i03. https://www.jstatsoft.org/article/view/v045i03
[10] “A systematic review of machine learning-based missing value imputation techniques,” Data Technologies and Applications, vol. 55, no. 4, pp. 558–585, 2021. DOI: 10.1108/DTA-12-2020-0298. https://doi.org/10.1108/DTA-12-2020-0298
[11] “Machine-learning Algorithms for Ischemic Heart Disease Prediction: A Systematic Review,” 2023. Free full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC10201879/

Gurram Sravan Kumar, Dr. Arathi Chitla (2026). A Source-Aware Data Audit and Exploratory Analysis Framework for a Hybrid Coronary Artery Disease Dataset. International Journal of Computer Techniques, 13(6), 638–646. ISSN: 2394-2231. DOI: https://doi.org/10.5281/zenodo.24128940