همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Resilience or Recession? A Longitudinal Analysis of NFT Market Dynamics and AI-Generated Value During the Crypto Downturn

نویسندگان :

Aria Shakoo¹ Amirhossein Ashrafian² Maedeh Mosharraf³

1- دانشگاه شهید بهشتی 2- دانشگاه شهید بهشتی 3- دانشگاه شهید بهشتی

کلمات کلیدی :

NFT،Crypto Winter،Generative AI،Price Prediction،Deep Learning،Random Forest،Financial Modeling

چکیده :

The Non-Fungible Token (NFT) market has transitioned from the speculative hyper-growth of 2021-2023 to a period of stabilization and correction in 2024-2025, often termed the “crypto winter.” While prior literature has extensively documented value drivers in bull markets, the asset pricing dynamics during prolonged market contractions remain underexplored. This paper presents a comprehensive longitudinal analysis of this downturn, focusing on two critical dimensions: the valuation reversal of Artificial Intelligence (AI)-generated assets versus traditional hand-drawn collections, and the comparative robustness of machine learning models in predicting prices under high-volatility conditions. We curated a dataset of over 113,000 transactions from 2024-2025 and benchmarked findings against a 2023 baseline dataset of near 7 million transactions. Our results reveal a stark inversion of market sentiment: while AI assets commanded a premium in 2023, they traded at a near 76% discount compared to traditional collections in 2024 (µ ≈ 0.92 Ether (ETH) vs 4.16 ETH), signaling a “flight to quality.” Furthermore, we evaluated two pricing models: a Deep Learning (DL) Multi-Layer Perceptron (MLP) and an ensemble Random Forest (RF). Contrary to the hypothesis that deep neural networks would better capture non-linear market volatility, the RF model (R 2 = 0.74) significantly outperformed the MLP (R 2 = 0.71) on the winter data. This comparative study suggests that in sparse, high-noise financial environments, ensemble decision trees remain more robust than deep architectures.