Introduction:

Bitcoin has been increasingly regarded as an investment asset.Bitcoin’s value reflects investors’ confidence in cryptocurrency.Because of its highly volatile nature, there is a need for good predictions on which to base investment decisions. Existing studies have leveraged machine learning for more accurate Bitcoin price prediction, few have focused on the feasibility of applying different modelling techniques to samples with different dimensional features.

Motivation:

Previous works simply concentrates only on accuracy when applying machine learning algorithms without considering the sample dimension.Applying machine learning without considering sample dimension causes overfitting problem.Stock market prediction using daily data and accessible high-frequency data has grown over decades.Predicting Bitcoin price in similar manner is lacking

Objective:

Leveraging appropriate Machine learning algorithms based upon characteristics of sample and dimension to predict Bitcoin price

Methodology:

Features for prediction models of Bitcoin daily price are : Block Size, Hash Rate, Mining Difficulty, Number of Transactions, Confirmed Transactions per Day, Mempool Transaction Count, Mempool Size, Total Transaction Fees, Market Capitalization, Estimated Transaction Value, Time between blocks, Trades per minute, Google trend search volume index, and Gold spot price.

Features for prediction models of 5-minute interval Bitcoin trading price are : Price, Trading Volume, Open, Close, High and Low points

After apply wrapper method (backward elimination ) on Bitcoin daily price data for features selection . Following 12 features are selected for prediction models :Block Size, Hash Rate, Mining Difficulty, Number of Transactions, Confirmed Transactions per Day, Mempool Size, Mempool Transaction Count,Market Capitalization, Estimated Transaction Value, Total Transaction Fees, Google Trend Search Volume Index, and Gold Spot Price

For Bitcoin 5-minute interval price prediction model, feature data for tick trading data are poor or unavailable for very small intervals. Therefore, this
paper consider the original features, which include Open, High, Close, Low, Price and Volume

Implementation:

Develops a binary classification algorithm to predict the sign change of Bitcoin price

Binary dependent variables denoted as y ∈ { -1,1}. Here y = -1 indicates that the Bitcoin price drops while y= 1 indicates that the Bitcoin price increases.

Machine Learning Techniques used:

Logistic Regression Linear Discriminant Analysis Random Forest Xgboost Quadratic Discriminant Analysis Support Vector Machine Decision Tree K-Nearest Neighbors

Result:

Logistic Regression and Linear Discriminant Analysis performs better for daily data with high-dimensional features

XGBoost outperform other machine learning models for high-frequency data