<p>In traditional satellite signal processing, cycle slip detection relies on a subjective threshold, susceptible to global navigation satellite system (GNSS) sampling rates and ionospheric changes. To address this challenge, this study introduces an oversampling-trained AdaBoost machine learning-based cycle slip detection method, oversampling AdaBoost cycle slip detection (OACD). This method integrates features of the geometry-free combination, Melbourne-Wubbena combination, signal-to-noise ratio, and elevation angle. It leverages the AdaBoost algorithm to meet the challenges of small-sample cycle-slip data. Experimental results demonstrate that the AdaBoost model achieved 100% accuracy on the training set and exhibited a low false alarm rate across different sampling rates. Notably, the model significantly reduced the false alarm rate by 28% in the southern region under active ionospheric conditions and by 3% in the quiet northern region. Additionally, the model exhibits spatial generalization capability, reducing the false alarm rate to 2% both within the modeled region and for stations 250 km outside the region. The study indicates that the OACD method has certain advantages in cycle slip detection, reducing false alarms and providing a new technical approach to enhance the accuracy and reliability of GNSS positioning.</p>