What does variance in a predictive model indicate?

Variance in a predictive model refers to the sensitivity of the model to fluctuations in the training data. When a model has high variance, it means that it captures noise in the training data instead of the underlying distribution. This can occur when the model is overly complex, leading it to make predictions that vary significantly with changes to the training dataset.

High variance typically results in the model performing well on the training data but poorly on new, unseen data, as it fails to generalize beyond its training examples. This is in contrast to bias, which relates to the error introduced by approximating a complicated problem with a simplified model.

Understanding variance is essential for building robust predictive models, as it helps identify the balance needed between underfitting and overfitting, guiding choices like model complexity and regularization techniques. By focusing on variance, one can work toward achieving a model that generalizes well to new data.