The effect of selected stock market indices and commodities on crypto asset prices: An empirical study based on machine learning

The study aims to predict the impact of the S&P 500 and KOSPI indices, as well as gold (XAU) and silver (XAG) prices, on the prices of Bitcoin (BTC) and Ethereum (ETH), the two assets with the highest market value in the crypto asset markets, using machine learning models. Additionally, the study compares the predictive performance of Decision Tree, Logistic Regression, Support Vector Machines (SMO), Artificial Neural Networks, IBk, K-Star, and LWL algorithms for both cryptocurrencies. The data used in the analyses was obtained from the Investing.com platform and includes daily closing values for the period between January 2, 2019, and December 30, 2024. The dataset was created using a total of 6 variables and 1,431 daily observations and contains 8,586 data points. The Java-based open-source WEKA software was used in the modeling process. Performance evaluation was conducted using metrics such as accuracy, precision, sensitivity, F-score, MAE, RMSE, and Kappa coefficient. The Decision Tree algorithm achieved 100% success in all metrics in Bitcoin prediction, demonstrating the highest performance, while both the Decision Tree and IBk algorithms stood out in ETH prediction. The K-Star algorithm showed strong performance only for ETH. The overall performance of other algorithms was relatively low. The findings reveal that algorithm selection and the structural characteristics of the target asset play a significant role in predicting cryptocurrency prices. The study contributes to the literature by comparatively demonstrating the effectiveness of machine learning algorithms in financial asset predictions.