Reproduce “NFT Appraisal Prediction: Utilizing Search Trends, Public Market Data, Linear Regression and Recurrent Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shrey Jain, Camille Bruckmann, Chase McDougall
This paper explores factors influencing NFT valuations and evaluates machine learning models for prediction, drawing on public market data, NFT metadata, and Google search trends. A linear regression model identified previous day's average price, daily trading volume, gas fees, and relative search volume as key predictors of NFT prices, achieving an adjusted R² of 0.65 on a test set, while recurrent neural networks were found unsuitable for direct absolute price forecasting in this volatile market.
Key insights
- Direct absolute price prediction in the highly volatile NFT market using RNNs (specifically GRU) proved unreliable, indicating the market's inherent complexity for straightforward forecasting.
- A linear regression model can effectively identify and quantify the influence of specific features on NFT valuations, even if direct price prediction is challenging.
- Public attention (Google search volume) and blockchain-specific metrics (daily trading volume, gas fees) are quantifiable factors that influence NFT prices.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.