Dynamic NFT Price Prediction: Incorporating Temporal Features into Machine Learning Models (BAYC Case Study)

Valuing non-fungible tokens (NFTs) challenging due to their unique nature and the extreme volatility of NFT markets. Conventional price-prediction approaches often rely on static analyses that fail to capture and model the temporal dependencies inherent in market behavior. This study addresses that research gap by introducing and evaluating dynamic models for NFT price prediction, using the Bored Ape Yacht Club (BAYC) collection as a case study. Leveraging a comprehensive dataset that combines intrinsic token attributes, rarity scores, trade volumes, and prior sale prices, together with cryptocurrency market indicators (Ethereum and Bitcoin) represented as temporal features, we develop machine-learning models that explicitly exploit temporal structures to forecast prices. Using time-series-based machine learning models, we capture sequential market patterns and conduct out-of-sample forecasts. Results show that the dynamic approach significantly outperforms static baselines, yielding prediction errors below ۱۰%. The findings underscore the necessity of incorporating temporal dynamics into NFT market analysis to achieve both higher predictive accuracy and deeper behavioral insights.

کلیدواژه ها:

NFT Price Prediction ، Dynamic Modeling ، Temporal Analysis ، Machine Learning ، Digital Asset Economics ، BAYC