CryptoPricePredictor

About

Project Proposal

Overall Objective: Predict future prices of various cryptocurrency

• Implement two machine learning algorithms
• Find datasets containing detailed history of various coins to train our algorithm
• Compare performance of different algorithms

Justification: Being able to accurately predict prices of cryptocurrency could enable someone to make smarter investments.

Algorithms of Choice

Time Series Analysis - Regression

• Support Vector Machine (SVM)

  • Widely applied and well surveyed
  • Non-linear and non-stationary process

• Generalized Additive Model (GAM)

  • Application: Facebook's Prophet, a forecasting tool
  • Decomposition: Trend + Cyclical + Seasonal + Irregular

Datasets

Kaggle Datasets

• Bitcoin's data at 1-minute inverals from Jan 2012 - Jan 2018

  • Attributes
    • Timestamp
    • Price (Open/High/Low/Close)
    • Volume in BTC & USD (Value of amount transacted in 24 hours)

• Several top coin's data at daily intervals over several years

  • Attributes
    • Timestamp
    • Price (Open/High/Low/Close)
    • Market Cap (Coin Price * Circulating Supply)

Authors

• Benjamin Carpenter
• Jacob Pauly
• Shuning Jin
• Tristan Larsin