kNN + Naive Bay

K-Nearest Neighbors

Known Nearest Nodes

Knowledge Network Navigator

Kernel Nearest Network

The number of clusters

The distance metric

The number of nearest neighbors to consider when making a prediction

The number of features in the dataset

Euclidean distance, Manhattan distance, or Minkowski distance

Correlation distance, Angular distance, or Time-series distance

Mahalanobis distance, Dice distance, or Edit distance

Logistic distance, Fisher distance, or Dynamic distance

Classification only

Regression only

Both, but it is most commonly used for classification

Neither classification nor regression

KNN becomes more accurate with large datasets

KNN loses its memory with large datasets

KNN is unable to handle large datasets due to overfitting

KNN can be computationally expensive and slow for large datasets

All features are dependent on each other

It assumes all features are independent of each other

It assumes all features have the same distribution

It assumes features are randomly distributed

Probabilistic algorithm

Deterministic algorithm

Cluster-based algorithm

Distance-based algorithm

• Pythagorean Theorem

Bayes' Theorem

Central Limit Theorem

Euclidean Theorem

•  They are calculated from the training data and represent the overall probability of a class.

• They are always set to be equal for all classes.

•  They represent the likelihood of the features given the class.

• They are only used when no training data is available.

•  Poorly, because it requires significant computation power for large datasets.

Well, because it is not affected by the curse of dimensionality.

Poorly, because it struggles with highly correlated features.

Well, because it ignores interactions between features.