A model of training data that, by taking too many of the data's quirks and outliers into account, is overly complicated and will not be as useful as it could be to find patterns in test data. See [..]

A machine learning problem whereby an algorithm is unable to discern information relevant to its assigned task from information which is irrelevant to its assigned task within training data. Overfitting therefore inhibits the algorithm’s predictive performance when dealing with new data.

Once you build a predictive model, it’s a lot more interesting to ask what it would predict in a new situation that’s never been seen before. It turns out that all algorithms, by their nature, perform better on data that they have seen before (known as in-sample or training data), versus data they have never seen before (known as out-of-sample data [..]

When attempting to fit a curve to a set of data points, producing a curve with high curvature that fits the data points well but does not model the underlying function well, its shape being distorted [..]