Classification of Financial Data Using Quantum Support Vector Machine

TL;DR

The paper investigates applying quantum support vector machines (QSVM) to financial data, addressing limitations of classical SVMs on non-stationary, high-dimensional stock market data. It evaluates multiple quantum feature maps, with the Pauli Y YY map delivering the strongest empirical advantage on the DSEx Broad Index dataset, validated against a classical SVM baseline and guided by the Phase Space Terrain Ruggedness Index (PTRI). A resource-estimation framework and hardware-in-the-loop experiments (IBM simulators and Nairobi QPU) demonstrate practical considerations for scaling. The study provides an end-to-end demonstration of quantum advantage in finance and outlines concrete directions for future work, including larger datasets and custom feature maps tailored to financial data characteristics.

Abstract

Quantum Support Vector Machine is a kernel-based approach to classification problems. We study the applicability of quantum kernels to financial data, specifically our self-curated Dhaka Stock Exchange (DSEx) Broad Index dataset. To the best of our knowledge, this is the very first systematic research work on this dataset on the application of quantum kernel. We report empirical quantum advantage in our work, using several quantum kernels and proposing the best one for this dataset while verifying the Phase Space Terrain Ruggedness Index metric. We estimate the resources needed to carry out these investigations on a larger scale for future practitioners.

Figures (6)

• Figure 1: Comparing average Balanced Accuracy of quantum kernels with classical SVM on datasets where classical SVM performance is closest to mean performance, across varying dataset sizes and 7 features.
• Figure 2: Cropped circuit of Pauli Y YY Feature Map for Quantum Kernel Calculation.
• Figure 3: Difference of average Balanced Accuracy of classical SVM vs Pauli Y YY Kernel on datasets where classical SVM performance is closest to minimum performance, across varying sizes and 5 to 7 number of features.
• Figure 4: Illustration of the Classical and Quantum classification approaches.
• Figure 5: PTRI scores for classical SVM and QSVM in a $15$ point configuration space.