Integrating Fuzzy Logic with Grey Relational Analysis for Enhanced Cryptocurrency Transaction Modelling

This study develops an integrated Fuzzy-Grey Relational Analysis (GRA) framework for modelling cryptocurrency transaction volume. Using daily Bitcoin data from July 2010 to March 2025 (n = 5359 observations), we perform GRA with sensitivity analysis (ρ = 0.1–0.9) to identify on-chain and market factors exhibiting the strongest geometric association with transaction volume. Network activity metrics—active addresses (GRG = 0.920), transaction count (GRG = 0.907), and miner revenue (GRG = 0.840)—are more closely related to volume than price (GRG = 0.821) or market sentiment (GRG = 0.673) across all market regimes. These three factors serve as inputs to a Mamdani-type Fuzzy Inference System with three empirically parameterized rules. In walk-forward validation (145 windows, 30-day horizon, 4350 out-of-sample predictions), the Fuzzy-GRA model reduces RMSE by 32.2% and sMAPE by 13.5% compared to linear regression. Ablation tests confirm that on-chain variables account for 89% of this predictive improvement. The Diebold-Mariano test verifies statistical significance (p < 0.0001), while overfitting diagnostics reveal better model generalization (overfit ratio: 0.081 vs. 0.126). The findings imply that transaction volume exhibits stronger geometric association with network activity metrics than with price or sentiment indicators, with implications for liquidity analysis and risk management.