IterativeStratificationSplitter<T>
Models & Types · AiDotNet.Preprocessing.DataPreparation.Splitting.Stratified
Iterative stratification splitter for multi-label classification problems.
Inheritance: DataSplitterBase<T> → IterativeStratificationSplitter<T>
For Beginners
Standard stratification works well for single-label classification where each sample belongs to exactly one class. But what about multi-label problems where each sample can have multiple labels (like a movie being both “Action” AND “Comedy”)?
How It Works
How It Works:
- Calculate label frequencies across all samples
- Start with rarest label combinations
- Iteratively assign samples to folds, prioritizing balance for rare labels
- Continue until all samples are assigned
Example: Document classification where each document can have multiple topics:
- Doc1: [Politics, Economy]
- Doc2: [Sports]
- Doc3: [Politics, Sports]
- Doc4: [Economy]
Constructors
| Constructor | Summary |
|---|---|
IterativeStratificationSplitter(double testSize, Matrix<T> labelMatrix, bool shuffle, int randomSeed) |
Creates a new Iterative Stratification splitter for multi-label data. |
Properties
| Property | Summary |
|---|---|
bool RequiresLabels { get; } |
|
string Description { get; } |
Methods
| Method | Summary |
|---|---|
DataSplitResult<T> Split(Matrix<T> X, Vector<T> y) |
|
IterativeStratificationSplitter<T> WithLabelMatrix(Matrix<T> labelMatrix) |
Sets the multi-label matrix. |