Cryptocurrency markets exhibit high levels of volatility and unpredictability. Determining which dimension reduction techniques are robust to anomalous market shocks is an important goal to construct investment portfolios with low risk. We considered principal component analysis (PCA), robust principal component analysis (RPCA), and autoencoders to estimate the intrinsic dimension of the 13 largest cryptocurrencies in terms of market capitalization. We found autoencoders performed the best of the three methods; however, they all suggested that most of the variance in the data can be explained with only one or two orthogonal variables, indicating that there exists a high level of interdependency between assets.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Rajotte, Benjamin
• Layne, Harrison
• Tavares, Taenler
• Peterson, Cole

Publisher

• Worcester Polytechnic Institute

Identifier

• 105871
• E-project-042723-102637

Keyword

• Cryptocurrency
• Dimension Estimation
• Intrinsic Dimension
• Neural Network

Advisor

• Wang, Fangfang
• Sturm, Stephan

Year

• 2023

Date created

• 2023-04-27

Resource type

• Major Qualifying Project

Major

• Data Science
• Mathematical Sciences

Source

• E-project-042723-102637

Rights statement

• In Copyright

Last modified

• 2023-06-18