Simple implementation of KNN and fuzzy KNN algorithms with a small class that compares their results with cross-validation across multiple k-values
Cross validation balances classes in data used for training, so that each class is present in number proportional to its occurences in full data set. Useful if data happens to be sorted by class.
Uses epsilon for 0-distance neighbors.
Supports euclidean and Manhattan distances.
Index.js has usage examples.