When Microseconds Decide Millions, Some Researchers Are Turning to Quantum Physics

In a new line of research, scientists are testing whether the strange rules of quantum mechanics can help financial systems see risk more clearly - and faster - than today's algorithms.

By the time a human trader blinks, thousands of financial transactions may already have occurred. In high‑frequency trading, where algorithms buy and sell assets in fractions of a second, speed is not a luxury - it is survival. Yet as markets become more interconnected and data grows more complex, even the most advanced classical machine‑learning systems are beginning to show strain.

That tension - between ever‑faster markets and the limits of conventional computation - has pushed some researchers to look in an unexpected direction: quantum physics.

In a recently published, open‑access research chapter, Rana Veer Samara Sihman Bharattej Rupavath and an international team of collaborators examined whether quantum machine learning, a field that blends artificial intelligence with quantum computing, could improve how financial systems detect and manage risk in high‑frequency trading environments. The work appears in the Proceedings of the International Conference on Sustainable Business Practices and Innovative Models (ICSBPIM‑2025), a peer‑reviewed venue that brings together research at the intersection of technology, business, and society -

https://www.atlantis-press.com/proceedings/icsbpim-25/126017615

Rana Veer Samara Sihman Bharattej Rupavath is affiliated with the Department of Business Administration at National Louis University in Tampa, where his research focuses on applying advanced computational methods - particularly machine learning and emerging quantum techniques - to real‑world business and risk‑management problems. His work frequently sits at the boundary between theoretical models and operational decision‑making, reflecting a broader trend in which business scholars increasingly engage with frontier technologies traditionally associated with physics and computer science.

At first glance, the idea sounds almost science‑fictional. Quantum computers rely on qubits that can exist in multiple states at once - a phenomenon known as superposition - and can become entangled in ways that defy everyday intuition. But these properties also allow quantum systems to represent and process information in extremely high‑dimensional spaces, something classical computers struggle to do efficiently.

Financial markets, it turns out, share a similar complexity.

"High‑frequency trading data is not just large - it is dense, nonlinear, and constantly changing," said researchers familiar with the study. "That combination makes it a natural candidate for experimentation with quantum‑inspired methods."

Testing Quantum Models on Market Risk

The research focused on two quantum machine‑learning approaches that have gained attention in recent years: Quantum Support Vector Machines (QSVM) and Variational Quantum Classifiers (VQC). Both are quantum analogues of familiar classification tools used to sort data into categories - such as "risk" and "no risk" - but they operate using quantum circuits rather than purely classical calculations.

QSVM uses what is known as a quantum kernel, mapping financial data into a quantum feature space where subtle patterns may become easier to separate. VQC, by contrast, relies on a parameterized quantum circuit that is iteratively adjusted by a classical optimizer, blending quantum computation with traditional machine‑learning techniques.

To test these models, the researchers worked with a dataset described as 39,221 high‑frequency trading samples, designed to reflect rapid market conditions. Before applying any quantum methods, they subjected the data to extensive preprocessing - removing statistical outliers, correcting for class imbalance using synthetic oversampling techniques, and standardizing features to avoid misleading results.

Those steps may sound mundane, but they are critical. In financial risk systems, false confidence can be dangerous. A model that appears accurate simply because it predicts "no risk" most of the time may fail catastrophically when real danger emerges.

Promising Results, With Caveats

According to the study's reported results, both quantum models performed strongly in classifying risk conditions. The QSVM model achieved 91.3 percent accuracy, while the VQC model reached 92.7 percent accuracy in the experiments described. Measures such as precision, recall, and F1‑score - metrics commonly used to evaluate classification reliability - were similarly high.

Perhaps more telling were the models' ROC‑AUC scores, which indicate how well a system distinguishes between classes across different thresholds. The study reports an AUC of 0.96 for QSVM and 0.98 for VQC, suggesting robust separation between risk and non‑risk scenarios under the tested conditions.

But accuracy alone is not the ultimate measure of success in financial systems. What matters just as much is how a model is wrong.

In a detailed error analysis, the researchers examined false positives - cases where low‑risk situations were flagged as dangerous - and false negatives, where genuine risks were missed. The latter can be especially costly in trading environments, potentially exposing firms to sudden losses. In the reported results, the VQC model showed fewer overall misclassifications and a more balanced error profile than QSVM, though both struggled with borderline cases that remain difficult for any algorithm to classify cleanly.

Not a Quantum Takeover - Yet

Despite the encouraging findings, the researchers are careful not to oversell their conclusions. Today's quantum hardware remains limited, with relatively few qubits and susceptibility to noise that can disrupt calculations. As a result, much of the work was conducted using quantum simulators rather than large‑scale, fault‑tolerant quantum machines.

The study emphasizes future directions rather than immediate deployment: improving quantum algorithms, integrating hybrid quantum‑classical systems, and testing models on more diverse and noisy financial data. These cautions echo broader assessments in the growing field of quantum finance, which suggest that practical advantages are likely to emerge incrementally rather than through sudden disruption.

A Glimpse of a Hybrid Future

For Rana Veer Samara Sihman Bharattej Rupavath and his co‑authors, the work represents a step toward that hybrid future: one in which classical and quantum systems collaborate rather than compete. By grounding their study in established evaluation practices - clear preprocessing steps, multiple performance metrics, and transparent discussion of limitations - the researchers aim to separate realistic progress from hype.

In an era when algorithms increasingly shape economic outcomes, even incremental improvements in understanding and managing risk can ripple outward. Whether quantum machine learning ultimately transforms finance or simply sharpens its analytical tools, the question may not be whether markets will adopt quantum ideas - but how quickly they can learn to trust them.