The Akaike Information Criterion (AIC) is a statistical measure used to compare different models and select the one that best balances goodness of fit with simplicity (i.e., a parsimonious model). A lower AIC value indicates a model that fits the data well without being overly complex. It penalizes models with more parameters to avoid overfitting.
For example, if a researcher builds two different regression models to predict house prices—one simple and one more complex—the model with the lower AIC would be preferred, even if the complex model fits slightly better, because the added complexity might not justify the improvement.
« Back to Glossary Index