To reduce trading risk for consumers in the Non-Fungible Token (NFT) market, characterized by high uncertainty, we develop an integrated approach for NFT feature evaluation and price prediction. Our work has several contributions. The first is to identify critical factors from different perspectives, including artwork, market, and social aspects. These features represent the intrinsic and extrinsic values of NFTs. The second is to adopt machine learning methods to establish relationships between variables and prices and then unify the features to construct models for price prediction. To evaluate our approach, we have conducted a series of experiments using real transaction data to investigate NFT trading across different market types. The results prove the effectiveness of our strategy across performance metrics of accuracy, mean-squared error, R 2 , and return of investment. Moreover, the analysis shows that the influence of the anchor price set by the artists and the reference price offered by previous purchasers is significant, suggesting that consumers may rely more on them when making purchasing decisions.
Chen et al. (Thu,) studied this question.