Publications Database

Welcome to the new Schulich Peer-Reviewed Publication Database!

The database is currently in beta-testing and will be updated with more features as time goes on. In the meantime, stakeholders are free to explore our faculty’s numerous works. The left-hand panel affords the ability to search by the following:

  • Faculty Member’s Name;
  • Area of Expertise;
  • Whether the Publication is Open-Access (free for public download);
  • Journal Name; and
  • Date Range.

At present, the database covers publications from 2012 to 2020, but will extend further back in the future. In addition to listing publications, the database includes two types of impact metrics: Altmetrics and Plum. The database will be updated annually with most recent publications from our faculty.

If you have any questions or input, please don’t hesitate to get in touch.

 

Search Results

Irene Henriques, Perry Sadorsky (2023). "Forecasting NFT Coin prices Using Machine Learning: Insights into Feature Significance and Portfolio Strategies", Global Finance Journal, 58, 100904.

View Paper

Abstract With the rise in popularity of Non-Fungible Tokens (NFTs), the demand for NFT coins has also surged. NFT coins are cryptocurrencies that facilitate NFT ecosystems by supporting NFT trading and platform governance. Accurate price predictions of NFT coins are crucial for risk managing volatility and constructing optimal portfolios. This study employs machine learning techniques to predict the daily price direction of four key NFT coins, namely ENJ, MANA, THETA, and XTZ. The machine learning methods employed include three decision tree-based methods (random forests, extremely randomized trees, XGBoost), support vector machine, Lasso and Naïve Bayes. The findings show that random forests, extremely randomized trees, XGBoost, and support vector machine models have accuracy ranging between 80% and 90% for predictions in the 14 to 21 day range. This adds to the literature showing that machine learning methods have high prediction accuracy for cryptocurrency prices. Conversely, Lasso or Naïve Bayes models yield considerably lower prediction accuracy. Feature importance is assessed using Shapley values. The Shapley value feature importance calculated from random forests highlights that, for 14 and 21-day forecasts, four variables - five-year expected inflation, ten-year bond yields, the interest rate spread, and on balance volume - are consistently highly ranked across all NFT coins. Additionally, the MA50, MA200, and WAD also emerge as important features. These results highlight the importance of including macroeconomic variables which capture business cycle conditions and technical analysis indicators that capture investor psychology as features. NFT coin portfolios constructed using trading signals generated from Extra Trees outperforms a buy and hold portfolio. Extra Trees are easy and fast to implement and investors not making use of this information are likely making sub-optimal investment decisions.