SMOTENCGenerator<T>

Models & Types · AiDotNet.NeuralNetworks.SyntheticData

SMOTE-NC generator that creates synthetic minority samples by interpolating between existing minority samples and their k-nearest neighbors, supporting both continuous and categorical features.

Inheritance: SyntheticTabularGeneratorBase<T> → SMOTENCGenerator<T>

For Beginners

SMOTE-NC generates new minority samples by “mixing” existing ones:

Given a minority sample and a nearby neighbor:

  • For numbers (age, income): pick a random point between the two values

Example: if sample has age=30 and neighbor has age=40, synthetic might have age=35

  • For categories (gender, region): use the most common value among the k neighbors

Example: if 3 of 5 neighbors have region=“East”, synthetic gets region=“East”

This is simpler and faster than GANs, and works well for structured tabular data.

How It Works

SMOTE-NC (Nominal and Continuous) operates as follows:

  1. Extract all minority class samples from the training data
  2. For each minority sample, find its k nearest neighbors among other minority samples
  3. Generate synthetic samples by interpolating between samples and randomly chosen neighbors

Constructors

Constructor Summary
SMOTENCGenerator(SMOTENCOptions<T> options) Initializes a new instance of the SMOTENCGenerator class.