DESIGN AND IMPLEMENTATION OF A NON-FUNGIBLE TOKEN PRICE PREDICTION SYSTEM USING MACHINE LEARNING

ABSTRACT This project was about building a system to predict NFT prices using Python. It combined different machine learning models to try and get good estimates based on data. I think the main goal was to make something practical for people dealing with NFTs. The setup used a Flask web app. That let users upload images of NFTs or put in metadata to get predictions. Each part of it handled preprocessing the inputs, pulling out features, and then running the models. The models included Linear Regression for a basic quick look, KNN which looks at similar nearby things, and LSTM for the time-based stuff like how prices change over time. They all gave back an estimated price plus some confidence score. It feels like mixing them helped balance things out, not too slow but still reliable enough. For the image side, it probably extracts visual features somehow. And for metadata, it uses attributes like rarity or whatever. I am not totally sure how deep the feature extraction goes, but it seems to work for both static info and the dynamic market shifts. Linear Regression gives that straightforward explanation, which is nice for understanding basics. KNN captures patterns from similar NFTs, I suppose. Then LSTM deals with sequences, like past prices influencing now. The web interface made it easy to use, just through a browser. No need for setting up Python or anything complicated. That is useful for casual users who collect NFTs or even investors who want quick checks. It supports both kinds of inputs, so pretty versatile. The whole thing is modular, which means it could grow. Like adding connections to blockchain for real data, or bigger datasets, maybe even fancier deep learning later on to boost accuracy. Some people might think simpler models are enough, but others want the full picture with time elements. Overall, this shows how machine learning pairs up with simple web tools to handle NFT values in real time. The digital asset world changes fast, and this kind of system fits in there somehow. Keywords: Non-Fungible Tokens (NFTs), K-Nearest Neighbors, Long Short-Term Memory, Flask Application, Digital Assets