Naive Bayes assumes that the presence of one feature does not affect the presence of another. What is this assumption called?

The assumption that the presence of one feature does not affect the presence of another is known as feature independence. In the context of the Naive Bayes classifier, this assumption simplifies the computation of probabilities for predictive modeling by treating features as if they are independent given the class label. This means that the classification of an instance is based solely on the individual contribution of each feature, rather than considering any interaction or correlation between them.

This assumption is particularly important because it allows Naive Bayes to work efficiently even with high-dimensional datasets, where features might otherwise interact in complex ways. While feature independence is often a simplification that doesn't hold in all real-world scenarios, Naive Bayes can still perform surprisingly well in practice due to this simplistic model structure, especially in text classification and spam detection tasks.

In contrast, feature selection refers to the process of selecting a subset of relevant features for use in model construction, while conditional dependence indicates a scenario where the presence of one feature does impact the presence of another. Feature correlation involves measuring how closely two features move with respect to one another. None of these concepts align with the specific assumption made in Naive Bayes regarding feature independence.