NFT floor price prediction using Covalent API data and time-series model, developed in Python's Jupyter Notebook
This project gives a daily floor price prediction of any Ethereum NFT. Users can input NFT contract address and get a price prediction for the future. As a demonstration example, the model is built using the daily data of Bored Ape Yacht Club from April 30th 2020 to November 24th 2021 downloaded from Covalent; then it predicts daily prices for the period from November 25th 2021 to January 15th 2022. The prediction period can be easily adjust to any day in the future from now. The prediction is a trend of where the average floor price will go in the specified future period. The model is automated and can be run for other NFTs through simply replacing the contract address. There is no need for manual model parameter tuning. It is a tool for NFT investors to quickly see where the price of a collection can go in the near future.
How it's made
The prediction model is built using Python in Jupyter Notebook. The data is from Covalent API. The model used in this project is the time-series ARIMA model. It has the ability to use past prices to predict future prices, using components such as autoregressive terms (past prices), nonseasonal differences of the past prices, and the lagged forecast errors. It's a very common model for predicting stock price, and it can be also predict NFT price!