Lift in machine learning refers to a metric used to evaluate the effectiveness of a predictive model, particularly in classification tasks. It measures how much better a model performs compared to random guessing. Specifically, lift quantifies the increase in the likelihood of a positive outcome when using the model versus not using it at all. A higher lift value indicates that the model is more effective at identifying true positives, making it a valuable tool for assessing model performance in applications such as marketing, fraud detection, and recommendation systems. Understanding lift helps in optimizing models for better decision-making.