Quantum Bayesian Networks for Machine Learning in Oil-Spill Detection

TL;DR

The paper tackles oil-spill detection from satellite imagery under imbalanced data and limited quantum hardware by introducing Quantum Bayesian Networks (QBNs) that fuse probabilistic reasoning with quantum state preparation. It demonstrates the standalone strength of QBNs in precision/recall and shows that hybrid QBNs combined with classical ML models can achieve near-optimal AUCs ($\approx 0.99$), albeit with AUC limitations when using QBNs alone due to imbalance. Experimental evaluation includes a 48-feature setup mapped to $N$-qubit circuits, experiments on a 127-qubit IBM device, and Qiskit-based simulations, highlighting both the potential and the hardware-induced constraints of quantum-assisted environmental monitoring. The work provides a practical pathway for quantum-enhanced anomaly detection in ocean surveillance and outlines future optimization needs for robust, scalable deployment.

Abstract

Quantum Machine Learning (QML) has shown promise in diverse applications such as environmental monitoring, healthcare diagnostics, and financial modeling. However, its practical implementation faces challenges, including limited quantum hardware and the complexity of integrating quantum algorithms with classical systems. One critical challenge is handling imbalanced datasets, where rare events are often misclassified due to skewed data distributions. Quantum Bayesian Networks (QBNs) address this issue by enhancing feature extraction and improving the classification of rare events such as oil spills. This paper introduces a Bayesian approach utilizing QBNs to classify satellite-derived imbalanced datasets, distinguishing ``oil-spill'' from ``non-spill'' regions. QBNs leverage probabilistic reasoning and quantum state preparation to integrate quantum enhancements into classical machine learning architectures. Our approach achieves a 0.99 AUC score, demonstrating its efficacy in anomaly detection and advancing precise environmental monitoring and management. While integration enhances classification performance, dataset-specific challenges require further optimization.

Figures (8)

• Figure 1: Overview of the simulation process that compares satellite imagery data to identify and classify non-spill and oil-spill regions. The simulation, performed on both QC devices and non-quantum devices, mimics oil-spill patterns, and these simulated images are cross-referenced with satellite data and extracted as valuable information for enhanced classification accuracy. The process focuses on refining the detection and differentiation between oil-spill-affected areas and clean regions for precise environmental analysis.
• Figure 2: A conceptual representation of our contributions for imbalanced classification, illustrating the development of QBNs algorithm, its integration with machine learning models, and their impact on anomaly classification.
• Figure 3: Structure of the Bayesian Networks for oil-spill and non-spill Regions. The diagram illustrates the Bayesian networks, where yellow nodes represent the oil-spill region $P_{\text{oil}}$ and blue nodes represent the non-spill region $P_{\text{non-spill}}$. The central node describes the conditional probability $\left(P_{\text{non-spill}}|P_{\text{oil-spill}}\right)$, indicating how the presence of oil influences the classification of non-spill patches. The accompanying table shows the conditional probabilities, where the values reflect the relationship between non-spill and oil-spill regions.
• Figure 4: Overall framework for QBNs classification. The framework starts with the input of imbalanced data, which is then processed by the QBNs using Qiskit for quantum state preparation and classification of probabilistic events. The quantum predictions are assessed for accuracy, and the resulting quantum features are used to train and test various classical ML models. The output performance metrics include accuracy, precision, recall, AUC, and F1 score.
• Figure 6: Experimental setup of the proposed approach, the imbalanced dataset is processed within the QBNs framework, implemented using Qiskit. The framework undergoes training on both a quantum simulator and IBM QPUs. Model performance is evaluated using standard binary classification metrics, including precision, recall, accuracy, and F1 score.