Enhancing Predictive Accuracy and Fairness in Healthcare Through Advanced Data Analysis

Nallagatla, Venkata Rakesh Chowdary (2025) Enhancing Predictive Accuracy and Fairness in Healthcare Through Advanced Data Analysis. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (4MB) | Preview
Preview	PDF (Configuration Manual) Download (767kB) | Preview

Abstract

Readmissions of diabetic patients are very critical both clinically and financially to the health care systems. This paper would present a fairness-conscious and interpretable machine learning pipeline to make readmission predictions based on harmonized Electronic Health Records (EHRs) between two open-source Kaggle datasets. It used extensive preprocessing, such as missing value imputation, categorical encoding, and a SMOTE-based approach to deal with class imbalance and was trained with a Transformer architecture in order to learn complex feature interactions. The result on the test set of the combination of transformer-generated embeddings with a Random Forest classifier was called an accuracy of 87.9 percent and an AUC-ROC of 0.79 and outperforming any of the benchmarks of the models such as XGBoost and Logistic Regression. The Rationality analysis revealed equal fairness in the prediction levels of all genders (DI = 0.78), and interpretability was supported with the application of SHAP and LIME to determine the influential predictors that included diabetes medication status, the number of lab procedures, and time in hospital. The comparison to the previous works on segmentation of organic compounds, proves predictive capabilities of the proposed pipeline comparable to or better than those of other state-of-the-art tools. These results highlight the usefulness of complementing AI fairness assessment with explainable AI to increase transparency, build clinical legitimacy, and facilitate a method to informed decisions in minimizing the readmission of diabetic patients. The methodology and findings present a scalable model of bringing the complex AI models into healthcare environments but maintaining both accuracy and an equal playing field of results.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Tomer, Vikas UNSPECIFIED
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence R Medicine > Healthcare Industry
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	26 Aug 2026 09:40
Last Modified:	26 Aug 2026 09:40
URI:	https://norma.ncirl.ie/id/eprint/9650

Actions (login required)

View Item