EXPLORING THE EFFECTIVENESS OF TRADITIONAL TIME SERIES MODELS AND MACHINE LEARNING ALGORITHMS FOR PRICE FORECASTING

Abstract

This work is a comprehensive analysis of time series forecasting of price variation of three important financial assets: Bitcoin (BTC), gold, and crude oil. In order to forecast future prices accurately, the study uses a combination of conventional time series models such as ARIMA and advanced time series machine learning algorithms. Analyses of historical price data are conducted in order to evaluate the model’s capacity to identify underlying trends and patterns. The study results indicate that the neural network model exhibits high accuracy in forecasting BTC and gold prices. This finding suggests that the neural network’s ability to capture complex relationships and nonlinearities within the data is advantageous for these assets. On the other hand, the LSTM model demonstrates superior forecasting accuracy for crude oil prices. This indicates that LSTM’s ability to handle time dependencies and long-term memory proves effective in capturing the dynamics of crude oil markets. Overall, the research highlights the importance of considering different time series models when forecasting prices in financial markets. The outcomes underscore the significance of utilizing neural network models, particularly for BTC and gold, as well as the value of LSTM models for accurate predictions in crude oil markets. The results add to the body of knowledge already available in time series analysis and offer information to investors, analysts, and policymakers who want to choose wisely when it comes to various asset classes.