Quantum Feature Optimization for Enhanced Clustering of Blockchain Transaction Data

TL;DR

The paper tackles clustering of high-dimensional, noisy blockchain transaction data. It systematically compares classical K-Means, a hybrid approach using quantum random features, and a fully quantum clustering method trained with SwAV loss. Key findings show that even depth-1 quantum feature maps substantially improve clustering quality, and end-to-end QNN training yields further gains. This work demonstrates practical quantum feature optimization as a scalable tool for robust blockchain analytics, with implications for anomaly detection and fraud analysis.

Abstract

Blockchain transaction data exhibits high dimensionality, noise, and intricate feature entanglement, presenting significant challenges for traditional clustering algorithms. In this study, we conduct a comparative analysis of three clustering approaches: (1) Classical K-Means Clustering, applied to pre-processed feature representations; (2) Hybrid Clustering, wherein classical features are enhanced with quantum random features extracted using randomly initialized quantum neural networks (QNNs); and (3) Fully Quantum Clustering, where a QNN is trained in a self-supervised manner leveraging a SwAV-based loss function to optimize the feature space for clustering directly. The proposed experimental framework systematically investigates the impact of quantum circuit depth and the number of learned prototypes, demonstrating that even shallow quantum circuits can effectively extract meaningful non-linear representations, significantly improving clustering performance.

Figures (6)

• Figure 1: Flowchart of the experimental design comparing three approaches: (1) Traditional K-Means, (2) Quantum Random Feature Extraction + K-Means, and (3) Fully Quantum Clustering.
• Figure 2: Random QNN for generating weights for feature transformation NN. This figure depicts a randomly initialized variational quantum circuit $V(\Theta)$, composed of single-qubit $R_y$ rotations and CNOT entanglement. The output distribution is used as quantum features.
• Figure 3: Trained QNN for generating weights for feature transformation NN. This figure illustrates a trained quantum circuit $V(\Theta)$, whose parameters are optimized via SwAV loss. The circuit uses the same gate structure as in Figure \ref{['fig:random_QNN']} but undergoes end-to-end training to enhance clustering performance.
• Figure 4: Silhouette Score vs. Quantum Depth
• Figure 5: Davies-Bouldin Index vs. Quantum Depth