This study aims to improve the accuracy of machine learning-financial market forecasting. Financial markets are notorious for being unpredictable and volatile, making predictions particularly difficult for traders or analysts. This study introduces an enhanced trend-following algorithm, that is combined with the Average True Range (ATR) indicator, and proposes new adaptive signal confirmation models. When combined, they produce more accurate buy and sell signals. This captures market trends with minimal false signals through the application of machine learning. This general framework improves prediction accuracy and the risk-adjusted returns drastically over the naive models proposed to date, thus providing a practical way for traders and institutions dealing with increasing financial sophistication. The study paves the way for further improvements in AI optimized finance and incorporation of machine learning into state-of-the-art trading systems.