In this second video recording of the NFT sales forecasting group project, we are presenting the datasets we collected from OpenSea API and Kaggle.
For now, we have a complete listing of the 9761 witches from the Crypto Coven collection, with prices and images which we encoded with base64. This is enough to start training our model on a single but complete collection.
For the next step we still have two options in mind: whether collect 4 more big and popular collections of 10k NFTs such as CryptoPunks and Bored Apes or instead collect smaller NFT collections like the 124 ones minted on the Solana blockchain. We are also discussing how to interpret the price (target) variable y since NFTs are bought in ETH whose value in USD is not stable.
We will probably end up using USD as for sales prediction value but keep in mind than intra-chain exchanges are done using ETH as reference.