What is the term used for the phenomenon where a machine learning model's performance deteriorates over time due to changes in the patterns of data?

The term for the phenomenon where a machine learning model's performance deteriorates over time due to changes in the patterns of data is termed "Model Drift." This occurs when the statistical properties of the target variable, which the model was trained on, shift over time, making the model less effective as it encounters new data that may not align with the historical patterns learned during training.

Model drift can manifest in various ways, such as changes in user behavior, seasonality effects, or evolving market trends. Therefore, keeping track of performance metrics over time is crucial, as it may indicate the need for model retraining or adjustment.

Other options do not accurately describe this phenomenon. Data leakage refers to the unintended inclusion of information in the training set that wouldn't be available in a real-world scenario, compromising the model's validity. Overfitting occurs when a model is too complex, capturing noise rather than the underlying pattern, leading to poor performance on new data. Underfitting, on the other hand, happens when a model is too simplistic to capture the underlying trend, resulting in poor performance in both training and test sets.