• Entry
  • Reader's guide
  • Entries A-Z
  • Subject index

Overfitting is a problem encountered in statistical modeling of data, where a model fits the data well because it has too many explanatory variables. Overfitting is undesirable because it produces arbitrary and spurious fits, and, even more importantly, because overfitted models do not generalize well to new data.

Overfitting is also commonly encountered in the field of machine learning, including learning by neural networks. In this context, a learner, such as a classifier or an estimator, is trained on an initial set of samples and later tested on a set of new samples. The learner is said to be overfitted if it is overly customized to the training samples and its performance varies substantially from one testing sample to the next.

Because the nature of the overfitting ...

Loading
  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles