원문정보
초록
영어
The NFT (Non-Fungible Token) market has grown rapidly, especially in the field of digital art and collectibles. However, significant price differences are often observed even among NFTs with similar traits, suggesting that valuation mechanisms involve more complex factors beyond rarity scores. This study examines the Azuki NFT collection to analyze how visual features, rarity attributes, and buyer behavior influence market prices. Metadata including traits, rarity scores, ownership information, and listing prices were collected and processed. By combining image embeddings with principal component analysis (PCA) and regression modeling, the study finds that both structured metadata and visual characteristics significantly impact pricing. These findings shed light on NFT pricing dynamics and suggest new approaches for data-driven asset valuation.
목차
1. Introduction
2. Research Background
3. Research Methodology
3.1. Data Collection
3.2. Feature Engineering
3.3. Model Construction and Analysis
3.4. User Behavior Analysis and Image Clustering
4. Results
4.1. Feature Importance Analysis (SHAP Results)
4.2. Analysis of User Behavior Patterns
4.3. Image Clustering Analysis
4.4. High-Value NFT Identification Model(XGBoost Classification)
5. Discussion
5.1. Market Pricing Significance of Image Features
5.2 The Driving Role of User Behavior in the High-End NFT Market
5.3. The Indicative Power of Visual Clustering on Pricing
5.4 Academic Contributions and Practical Implications
5.5 Limitations and Future Research Directions
6. Conclusion
References