What aspect of a model does feature importance measure?

Feature importance specifically quantifies the contribution of each feature in a dataset to the predictions made by a machine learning model. This aspect is crucial because it helps in understanding which variables are most influential in the decision-making process of the model. By evaluating feature importance, data scientists can prioritize features that meaningfully impact the model's predictions and possibly discard those that are less important, leading to more efficient and interpretable models.

The other aspects mentioned—such as model complexity, overall accuracy, and duration of model training—do not directly relate to how individual features contribute to predictions. Model complexity pertains to the structure and learning capabilities of the algorithm used, accuracy reflects how well the model performs on its predictions in terms of correctness, and training duration relates to the time it takes to build the model based on the data provided. Understanding feature importance enriches one's ability to make informed decisions regarding feature selection and model optimization.