Deep learning predictive models for non-fungible tokens

Details

Supervisors

Hafner, Christian

Faculty

Faculté des sciences

Degree label

Master [120] en science des données, orientation statistique, à finalité spécialisée

Abstract

A non fungible token (NFT) is a digital identification of a certain asset. This new technology derived from the cryptocurrencies was launched in early 2017. It gives access to functionalities such as identification of digital art piece, valuation of digital collectibles, the avoid the copy-paste principal without giving credits, etc. This market became very influent in the last years. For instance, the number of dollars generated by the NFT market in 2021 amounts to 24.9 billion US dollars. This thesis will deeply analyze the market for different NFTs types and demystify all the growing interest about this new technology that can look disturbing at first sight. The main purpose of this academic paper is to build some deep learning models to predict the market from internal view such as NFT features and external view like social network influence. This paper is divided in four distinct chapters : 1. The first chapter describes the state of the art in the field of NFT market prediction. This section gives a preview of all the work previously done by the academics. 2. The second chapter will give a detailed definition of an NFT, how it is created, on which technology it relies. This will imply resources that have to be introduced, such as Ethereum blockchain, Smart contract, ERC-20 and ERC-721 protocols. An NFT can be classified in different classes called types. Each of them will be explained in detail in order to produce analyses and see the evolution of the market for every type. 3. The third chapter speaks about the data collection and the analysis. For each type of NFT, a descriptive analysis will be carried out. Furthermore, a global analysis of the market will be visualized. 4. The last chapter will present the elaboration of multiple predictive models in order to have an idea of the future of the NFT market. These models will be computed in different ways : a prediction of the overall market, a prediction of each collection individually and a prediction to a global model taking into account all the collections and the sentiments of the social media Twitter. The impact of social media on the NFT market will be explained in details.