Data Leakage

ˈdeɪtə ˈliːkɪdʒ

Data leakage refers to the unintentional exposure of information from the training dataset to the model during the training process, leading to overly optimistic performance metrics. It occurs when the model has access to data it should not have during training, which can include data from the test set or future data that would not be available in real-world scenarios. This can result in a model that performs well on the training data but fails to generalize to unseen data, ultimately compromising its effectiveness in practical applications. Common use cases for understanding data leakage include model development in fields such as finance, healthcare, and marketing, where accurate predictions are critical.