近年來由於虛擬化的潮流，帶動Non-Fungible Token(NFT)數位術品市場興起，且由於NFT市場尚未被大眾所熟悉，其價格的決定性因素與特性仍在探索階段。自NFT的交易資訊、市場特性到影像特徵、文字分析，近年各項研究提出由不同面向探討組成成交價格的因素與特性。
於2021年Nadini學者等人[1]提出以交易網絡與、相同系列作品的平均歷史交易價格與影像特徵幫助預測NFT價格的方法，當中以歷史交易價格對價格變異的解釋力最為重要，其他因素亦能幫助提升線性迴歸模型的解釋能力。並於隔年，2022年由Horky學者等人[2]提出將用於分析流動性低的房地產交易市場的Hedonic Regression，應用在具有相同特性的NFT數位藝術市場中，該方法同時考慮市場性因素、物件本身的特性與時間的影響。以及Wang學者[3]於2022年提出NFT中存在錨定效果，人們會依據一基準，錨定，漸進式調整價格的預測，並由不同的起始價格將會導致不同的結果。
綜上所述，本研究提出:基於機器學習模型之NFT價格預測模型與錨定效果。其方法使用線性與機器學習迴歸模型，同時考慮市場性因素、交易網絡因素、NFT本身的影像特性與歷史交易價格，並訂定NFT的錨定價格，有效提高NFT在初級與次級交易市場中，價格變異的解釋力。
最後實驗結果中，透過依序移除變數以驗證各項變數與價格存在線性或非線性關係，並以歷史交易價格與錨定價格對價格變異的解釋力影響最為明顯。

In recent years, the virtualization trend has driven the rise of the Non-Fungible Token (NFT) digital art market. However, the factors and characteristics influencing NFT prices are still being explored. Several studies have investigated transaction information, market characteristics, image features, and text analysis to understand NFT price determinants.
Nadini et al. [1] proposed a method using transaction networks, average historical prices of similar artworks, and image features to predict NFT prices. Historical prices were the most critical factor, while Horky et al. [2] applied Hedonic Regression to consider market factors, artwork characteristics, and time. Wang [3] identified an anchoring effect, with price predictions anchored to a reference point.
This study proposes a machine learning-based NFT price prediction model considering market factors, transaction networks, image features, and historical prices. Anchoring prices enhance the model's explanatory power. Experimental results show the significant influence of historical and anchoring prices on price variations.

論文審定書i
誌謝ii
摘要iii
Abstractiv
目錄v
圖次viii
表次viii
第一章緒論1
1.1.研究背景1
1.2.研究動機1
1.3.研究目的2
第二章文獻探討3
2.1.傳統藝術品定價3
2.2.錨定效果(Anchoring Effect)3
2.3.Non-Fungible Tokens(NFT) 視覺特徵5
2.3.1.使用AlexNet模型5
2.3.2.使用ResNet101模型6
2.4.影像辨識模型6
2.4.1.ResNet6
2.4.2.SlowFast7
2.4.3.X3D7
2.5.Hedonic Regression8
2.5.1.南非藝術品市場實證研究8
2.5.2.NFT市場應用研究9
2.6.機器學習迴歸模型10
2.6.1.Random Forest Regressor10
2.6.2.Gradient Boosting Regressor10
2.7.賽局理論(Game Theory)10
2.7.1.英式拍賣(English Auction)11
2.7.2.荷蘭式拍賣(Dutch Auction)11
2.7.3.貝氏賽局(Bayesian Game)12
第三章研究方法13
3.1.模型架構13
3.2.資料前處理15
3.3.預訓練影像模型16
3.3.1.ResNet5016
3.3.2.X3D16
3.4.損失函數17
3.5.微調(Fine-tuning)17
3.6.迴歸模型18
3.6.1.線性迴歸模型(Linear Regression)18
3.6.2.Random Forest Regressor18
3.6.3.Gradient Boosting Regressor18
第四章實驗20
4.1.資料集介紹20
4.2.實驗設計21
4.3.評估方式22
4.3.1.方根誤差(Mean Square Error, MSE)23
4.3.2.決定係數(Coefficient of Determination, R2)23
4.3.3.投資報酬率(Return on Investment, ROI)23
4.4.影像模型23
4.4.1.ResNet5023
4.4.2.X3D-S25
4.5.簡單線性迴歸 – JPEG & PNG25
4.6.Random Forest Regressor - JPG & JPEG29
4.7.Gradient Boosting Regressor – JPG & JPEG31
4.8.簡單線性迴歸 – GIF34
4.9.Random Forest Regressor – GIF35
4.10.Gradient Boosting Regressor – GIF38
4.11.投資報酬率檢驗40
4.12.結果分析41
第五章結論46
參考文獻47
附錄52

[1]Nadini, M., Alessandretti, L., Di Giacinto, F., Martino, M., Aiello, L. M., & Baronchelli, A. (2021). Mapping the NFT revolution: market trends, trade networks, and visual features. Sci Rep, 11(1), 20902. https://doi.org/10.1038/s41598-021-00053-8
[2]Horky, F., Rachel, C., & Fidrmuc, J. (2022). Price determinants of non-fungible tokens in the digital art market. Finance Research Letters, 48. https://doi.org/10.1016/j.frl.2022.103007
[3]Wang, Y. (2022). Anchoring Effect and Loss Aversion: Evidence from the Non-Fungible Token Market. Available at SSRN 4097185.
[4]OpenSea. https://opensea.io/
[5]SuperRare. https://superrare.com/
[6]Ante, L. (2021). Non-fungible token (NFT) markets on the Ethereum blockchain: Temporal development, cointegration and interrelations. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3904683
[7]White, B., Mahanti, A., & Passi, K. (2022). Characterizing the OpenSea NFT Marketplace Companion Proceedings of the Web Conference 2022, https://doi.org/10.1145/3487553.3524629
[8]Wang, Q., Li, R., Wang, Q., & Chen, S. (2021). Non-fungible token (NFT): Overview, evaluation, opportunities and challenges. arXiv preprint arXiv:2105.07447. https://doi.org/https://doi.org/10.48550/arXiv.2105.07447
[9]Park, A., Kietzmann, J., Pitt, L., & Dabirian, A. (2022). The Evolution of Nonfungible Tokens: Complexity and Novelty of NFT Use-Cases. IT Professional, 24(1), 9-14. https://doi.org/10.1109/mitp.2021.3136055
[10]Mekacher, A., Bracci, A., Nadini, M., Martino, M., Alessandretti, L., Aiello, L. M., & Baronchelli, A. (2022). How rarity shapes the NFT market. arXiv preprint arXiv:2204.10243. https://doi.org/https://doi.org/10.48550/arXiv.2204.10243
[11]Fazli, M., Owfi, A., & Taesiri, M. R. (2021). Under the skin of foundation nft auctions. arXiv preprint arXiv:2109.12321. https://doi.org/https://doi.org/10.48550/arXiv.2109.12321
[12]Piyadigama, D., & Poravi, G. (2022). An Analysis of the Features Considerable for NFT Recommendations. arXiv preprint arXiv:2205.00456. https://doi.org/https://doi.org/10.48550/arXiv.2205.00456
[13]Fraiberger, S. P., Sinatra, R., Resch, M., Riedl, C., & Barabási, A.-L. (2018). Quantifying reputation and success in art. Science, 362(6416), 825-829. https://doi.org/10.1126/science.aau7224
[14]McAndrew, C. (2010). Fine art and high finance: expert advice on the economics of ownership (Vol. 36). John Wiley & Sons.
[15]Ursprung, H. W., & Wiermann, C. (2011). Reputation, price, and death: An empirical analysis of art price formation. Economic Inquiry, 49(3), 697-715. https://doi.org/ https://doi.org/10.1111/j.1465-7295.2009.00279.x
[16]Hofstetter, R., de Bellis, E., Brandes, L., Clegg, M., Lamberton, C., Reibstein, D., Rohlfsen, F., Schmitt, B. H., & Zhang, Z. J. (2022). Crypto-Marketing: How Non-Fungible Tokens (NFTs) Challenge Traditional Marketing. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4055610
[17]Dowling, M. (2022). Is non-fungible token pricing driven by cryptocurrencies? Finance Research Letters, 44. https://doi.org/10.1016/j.frl.2021.102097
[18]He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition,
[19]Feichtenhofer, C., Fan, H., Malik, J., & He, K. (2019). Slowfast networks for video recognition. Proceedings of the IEEE/CVF international conference on computer vision,
[20]Feichtenhofer, C. (2020). X3d: Expanding architectures for efficient video recognition. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition,
[21]Chanel, O., Gérard-Varet, L.-A., & Ginsburgh, V. (1996). The relevance of hedonic price indices. Journal of Cultural Economics, 20(1), 1-24. https://doi.org/10.1007/s10824-005-1024-3
[22]Fedderke, J. W., & Li, K. (2020). Art in Africa: Hedonic price analysis of the South African fine art auction market, 2009–2014. Economic Modelling, 84, 88-101. https://doi.org/10.1016/j.econmod.2019.03.011
[23]Breiman, L. (2001). Random forests. Machine learning, 45, 5-32. https://doi.org/https://doi.org/10.1023/A:1010933404324
[24]Friedman, J. H. (2001). Greedy function approximation: a gradient boosting machine. Annals of statistics, 1189-1232. http://www.jstor.org/stable/2699986
[25]Osborne, M. J. (2004). An introduction to game theory. Oxford University Press.
[26]SuperRare Dataset. https://www.kaggle.com/datasets/franceschet/superrare
[27]Alon, I., Bretas, V. P. G., & Katrih, V. (2023). Predictors of NFT Prices. Journal of Global Information Management, 31(1), 1-18. https://doi.org/10.4018/jgim.317097
[28]Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database. 2009 IEEE conference on computer vision and pattern recognition,
[29]Kay, W., Carreira, J., Simonyan, K., Zhang, B., Hillier, C., Vijayanarasimhan, S., Viola, F., Green, T., Back, T., & Natsev, P. (2017). The kinetics human action video dataset. arXiv preprint arXiv:1705.06950.
[30]avant-galerie.com. https://avant-galerie.com/en/artists/robbie-barrat
[31]aiartists.org. https://aiartists.org/robbie-barrat
[32]artjaws.com. https://www.artjaws.com/en/portfolio/ronan-barrot-robbie-barrat/
[33]androidjones.com. https://androidjones.com/
[34]edm.com. https://edm.com/interviews/android-jones-psychedelics-visionary-arts