The Effect of Linear Discriminant Analysis and Quantum Feature Maps on QSVM Performance for Obesity Diagnosis

Abstract

Obesity, characterized by an excessive accumulation of body fat, is a significant health issue that predisposes individuals to numerous diseases. Therefore, early intervention and necessary measures in the diagnosis and treatment of obesity are of great importance. In medicine, classical machine learning algorithms are widely used to accelerate the prediction process. However, the increasing volume of data often renders these algorithms insufficient for accurate disease diagnosis. At this point, quantum computing-based algorithms offer more efficient and faster solutions by leveraging quantum physics, which operates contrary to the principles of classical physics. Dimensionality reduction techniques play a critical role in both classical and quantum classifiers. In this study, classical dimensionality reduction methods, namely Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA), were applied to an obesity dataset. The dataset was subsequently analyzed using Quantum Support Vector Machine (QSVM) and Support Vector Machine (SVM) algorithms. As part of the QSVM studies, three different quantum feature maps, which facilitate the quantum bit transformation of classical bit data, were also compared. The analysis revealed that the proposed LDA-QSVM method achieved 100% accuracy when used with the Z and Pauli X feature maps. This remarkable success, which is rarely seen in the literature on obesity data, underscores the potential of quantum-based algorithms in the diagnosis and treatment of obesity.