OPTIMIZING HEALTHCARE WITH MACHINE LEARNING : PROTECTING FINANCES AND IMPROVING DIAGNOSIS

Abstract

This thesis addresses critical challenges in healthcare through two distinct yet interrelated studies. The first paper tackles the escalating issue of financial fraud detection within healthcare systems, a pressing concern exacerbated by advancements in electronic payment methods. Traditional fraud detection approaches have proven inadequate, necessitating the development of novel solutions. This study introduces an ensemble fraud detection classifier, leveraging a combination of machine learning algorithms to enhance performance. Methodologically, the ensemble classifier undergoes rigorous evaluation utilizing accuracy, precision, and recall metrics, showcasing its superiority over conventional methods such as Naive Bayes, Random Forest, and K-Nearest Neighbours. With an accuracy of 99.46%, precision of 98.38%, and recall of 98.58%, the ensemble method significantly outperforms its counterparts, offering promising avenues for future research. Further investigations aim to integrate hybrid models tailored to address dataset imbalances and ensure real-time responsiveness in financial transactions. The second paper addresses the urgent need for rapid and accurate diagnosis of pneumonia from chest X-ray (CXR) images, a critical aspect of medical diagnostics with profound implications for patient care. Leveraging the Swin Transformer V2, an innovative deep learning architecture, this study explores its application to pneumonia diagnosis within the medical imaging domain. Methodologically, the study evaluates the model's performance against a diverse CXR dataset, including various conditions and manifestations of pneumonia. Comparative analysis with established deep learning architectures such as AlexNet, MobileNetV3, VGG-16, ResNet 50, and DenseNet highlights the Swin Transformer V2's superiority in identifying subtle patterns indicative of pneumonia, achieving an accuracy of 98.6%. The findings underscore the transformative potential of integrating advanced deep learning models into clinical diagnostic processes, offering unprecedented accuracy and paving the way for significant advancements in healthcare practices. Possible applications of this research include the integration of advanced diagnostic models into clinical settings, potentially revolutionizing healthcare practices. Future research directions may include exploring hybrid models combining deep learning with traditional diagnostic methods and optimizing models for real-time application in clinical settings.