Using Prophet for Accurate Time-Series Predictions of Doge Coin

Abstract

Cryptocurrencies, including Dogecoin (DOGE), exhibit extreme price volatility and speculative behavior, making accurate price prediction a significant challenge for traders and analysts. This study applies Facebook Prophet, a robust time-series forecasting model, to predict Dogecoin's price movements using historical price and trading volume data. Prophet's ability to handle irregular datasets, missing values, and complex seasonality makes it well-suited for volatile financial markets. The methodology includes preprocessing the dataset, training Prophet on the “Close” price, and evaluating its predictive performance using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The results reveal Prophet's capability to capture Dogecoin's underlying trends and seasonality, providing actionable insights into market behavior. By comparing Prophet's performance with traditional models like ARIMA and advanced deep learning techniques such as LSTM, the study underscores its strengths and limitations in cryptocurrency forecasting, contributing to the growing research on cryptocurrency analytics and offering a reliable framework for understanding and predicting price dynamics in highly volatile markets like Dogecoin.