Comparative Study of DEFI Lending Protocols using LSTM

Abstract

Decentralized finance is a revolutionary change in the financial system, using blockchain technology to build a diverse and open network of financial services. By cutting out middlemen, DeFi promotes financial equality and reaches out to more people. Smart contracts can complete transactions without the need for traditional banks to lower costs and improve efficiency in lending, borrowing, trading, and yield farming activities. The proposed research compares four leading DeFi lending protocols: AAVE, MAKERDAO, COMPOUND, and VENUS Finance. We have used Long Short-Term Memory (LSTM) neural networks to analyze historical data and measure key parameters, such as lending and borrowing rates, Total Value Locked (TVL), Market Capitalization, and token price dynamics. We found that AAVE and COMPOUND exhibit similar mean rates but AAVE offers more precise predictions. MAKER provides potentially higher returns but with a higher degree of unpredictability. VENUS, despite its precise predictions, yields the lowest returns due to its lower mean lending rate. Overall, the approach enhances the understanding of the dynamics within the DeFi ecosystem, helping stakeholders to make informed decisions. Index Terms—Decentralized, Blockchain, LSTM, Time Series Analysis.