Ethereum Price Prediction Machine Learning Model
Introduction:
The Ethereum Price Prediction Model is designed to forecast the future price of Ethereum using historical data spanning from 2015 to 2023. This project involves building and fine-tuning various machine learning models to predict Ethereum's price based on key features such as trading volume, market cap, and time-related variables. The aim of this project is to leverage advanced regression techniques to achieve accurate and robust price predictions for Ethereum, benefiting investors and financial analysts.
Skills and Libraries Used:
Python: Data manipulation, feature engineering, and model building.
Pandas: For data preprocessing, feature extraction, and handling time-series data.
NumPy: Mathematical operations and data manipulation.
Matplotlib & Seaborn: Data visualization and exploratory data analysis (EDA).
Scikit-learn: Implementation of machine learning models such as Linear Regression, Ridge Regression, Lasso, Decision Tree, Random Forest, and Support Vector Regression.
Statsmodels: Time series decomposition for trend and seasonality analysis.
Gradient Boosting Regressor: An ensemble learning method for improved predictive performance.
Cross-validation: To assess model performance on unseen data.
RandomizedSearchCV: Hyperparameter optimization to fine-tune the Random Forest model.
Results:
Several machine learning models were tested to predict Ethereum prices, with Random Forest and Gradient Boosting performing best. The Random Forest model achieved an RMSE of 14.14, outperforming other models like Decision Tree (RMSE: 17.15) and Gradient Boosting (RMSE: 15.29). Additionally, the Random Forest model provided a custom accuracy of 70.83% within a 1% threshold. The hyperparameter tuning of the Random Forest model using RandomizedSearchCV further optimized the performance by selecting the best parameters.
This model has the potential to serve as a reliable forecasting tool for Ethereum price prediction, with visualizations and performance metrics showing strong alignment between predicted and actual values.