Algorithmic Trading Using Technical Indicators and eXtereme Gradient Boosting

The essence of profitable trading is navigating the perilous balance between minimising risk and maximising consistent returns. Success hinges on precise timing in choosing the suitable buy and sell signals. This paper proposes a novel approach that treats this decision-making process as a classification problem. We introduce a unique method for labelling training data, enabling an XGBClassifier to glean accurate buy and sell signals from historical market data using technical indicators as input features. This data-driven strategy’s efficacy is tested across four diverse asset classes, encompassing the three largest US market cap sectors and the Cryptocurrency market. Our experimental results demonstrate that the proposed model generates reasonable returns, consistently outperforming a passive buy-and-hold strategy while considerably mitigating risk.