Research on Defi Lending Forecasting based on GRA and Rolling GM(1,N) Model

Abstract

This paper investigates the short-term dynamics of decentralised lending conditions on Aave, a leading DeFi pooled lending protocol whose rapid TVL growth has made its variable interest-rate mechanism systemically relevant for on-chain investors and risk managers. In contrast to traditional fixed-income markets, DeFi interest rates are endogenously determined by pool utilisation and are observed over relatively short and highly volatile time series, which poses challenges for conventional econometric models. To address this, we develop a grey-system forecasting framework tailored to the USDC pool on Aave v3. Weekly data from January to November 2025 are collected on the utilisation rate (U), total value locked (TVL) and the Crypto Fear & Greed Index (FG). Grey relational analysis (GRA) is first employed to screen explanatory factors; both TVL and FG exhibit relational grades above 0.7 with respect to U (0.7138 and 0.7673, respectively), confirming that on-chain liquidity and market sentiment jointly drive utilisation. On this basis, a multivariate rolling GM(1,3) model is specified, estimated over an 8-week moving window and used to generate one-step-ahead forecasts of weekly utilisation. The resulting pseudo out-of-sample forecasts for 46 weeks achieve a mean absolute percentage error of 5.97%, with a posterior error ratio C=0.6475 and small error probability P=0.7632, indicating acceptable forecasting accuracy under standard grey-model criteria. The model successfully reproduces the medium-term high-utilisation regime (around 0.8–0.88) that places the USDC pool near the kink of Aave’s dual-slope interest-rate curve, implying frequent transitions into a high-rate environment for borrowers while offering attractive yields to liquidity suppliers. By linking grey-based utilisation forecasts to the protocol’s endogenous rate formula, the paper provides an interpretable and data-efficient tool for anticipating Aave’s interest-rate conditions and supporting lending, borrowing and arbitrage decisions in DeFi.