Comparative Analysis of Statistical and Machine Learning Models for Enhanced Cryptocurrency Price Prediction with Technical Indicator Integration

Introduction

This repository contains the code and resources for a class project on forecasting cryptocurrency prices using statistical models and machine learning algorithms. The project explores the performance of Linear Regression, ARIMAX, Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), Long Short-Term Memory (LSTM) networks, Vector Autoregression (VAR), XGBoost, and LightGBM in forecasting the prices of Bitcoin, Ethereum, and Dogecoin.

Dataset

The dataset used in this project is collected from finance.yahoo.com, containing daily data for Bitcoin, Ethereum, and Dogecoin from 2018-03-01 to 2024-06-01. The dataset includes features such as open price, high price, low price, close price, adjusted close price, and trading volume.

Feature Engineering

To improve the predictive performance of the models, the project incorporates technical indicators as features. These indicators, such as moving averages and momentum indicators, provide insights into market sentiment and trends, aiding in the prediction of price movements.

Models

The following models are implemented and evaluated in this project:

• Linear Regression
• ARIMAX
• Vector Autoregression (VAR)
• Recurrent Neural Networks (RNN)
• Gated Recurrent Units (GRU)
• Long Short-Term Memory (LSTM)
• XGBoost
• LightGBM

Evaluation

The models are evaluated using various performance metrics, such as Root Mean Squared Error (RMSE), Mean Absolute Percent Error (MAPE), and Symmetric Mean Absolute Percentage Error (SMAPE). Three different train-test split ratios are used for each model: 70:30, 80:20, and 90:10. The results are compared to assess the strengths and weaknesses of each model in accurately predicting cryptocurrency prices.

Division of Work

Member	Responsibilities
Le Gia Kiet	Implemented ARIMAX, GRU, and LightGBM models. Contributed to data preprocessing and feature engineering.
Le Quoc Khanh	Implemented Linear Regression, LSTM, and XGBoost models. Assisted in model evaluation and result analysis.
Nguyen Thi Thuy	Implemented VAR and RNN models. Contributed to literature review and project documentation.

Contributing

Contributions to this project are welcome. If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.

Acknowledgments

We would like to thank our instructor and the University of Information Technology for providing the resources and support for this class project.