期刊
JOURNAL OF RISK AND FINANCIAL MANAGEMENT
卷 14, 期 5, 页码 -出版社
MDPI
DOI: 10.3390/jrfm14050198
关键词
gold and silver prices; forecasting; machine learning; random forests; bagging; stochastic gradient boosting
资金
- Schulich School of Business
This research uses machine learning methods to predict the price direction of gold and silver exchange traded funds, showing that tree bagging, stochastic gradient boosting, and random forests have higher accuracy rates compared to logit models. Stochastic gradient boosting has slightly lower accuracy than random forests for forecast horizons over 10 days.
Gold is often used by investors as a hedge against inflation or adverse economic times. Consequently, it is important for investors to have accurate forecasts of gold prices. This paper uses several machine learning tree-based classifiers (bagging, stochastic gradient boosting, random forests) to predict the price direction of gold and silver exchange traded funds. Decision tree bagging, stochastic gradient boosting, and random forests predictions of gold and silver price direction are much more accurate than those obtained from logit models. For a 20-day forecast horizon, tree bagging, stochastic gradient boosting, and random forests produce accuracy rates of between 85% and 90% while logit models produce accuracy rates of between 55% and 60%. Stochastic gradient boosting accuracy is a few percentage points less than that of random forests for forecast horizons over 10 days. For those looking to forecast the direction of gold and silver prices, tree bagging and random forests offer an attractive combination of accuracy and ease of estimation. For each of gold and silver, a portfolio based on the random forests price direction forecasts outperformed a buy and hold portfolio.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据