CombinatorialPurgedSplitter<T>
Models & Types ยท AiDotNet.Preprocessing.DataPreparation.Splitting.TimeSeries
Combinatorial Purged Cross-Validation splitter for time series data.
Inheritance: DataSplitterBase<T> โ CombinatorialPurgedSplitter<T>
For Beginners
This is an advanced cross-validation method designed specifically for financial time series where data leakage can invalidate backtests.
How It Works
How It Works:
- Divide data into n groups based on time periods
- Generate all combinations of k groups for testing
- For each combination, remaining groups form training set
- Apply purging (remove samples near test boundaries) and embargo (block after test)
Use Cases:
- Financial backtesting where overlapping data causes leakage
- Time series with autocorrelation that persists across samples
- Generating many independent train/test splits from limited time series
Constructors
| Constructor | Summary |
|---|---|
CombinatorialPurgedSplitter(int nGroups, int nTestGroups, int purgeSize, int embargoSize, int randomSeed) |
Creates a new Combinatorial Purged CV splitter. |
Properties
| Property | Summary |
|---|---|
int NumSplits { get; } |
|
string Description { get; } |
Methods
| Method | Summary |
|---|---|
DataSplitResult<T> Split(Matrix<T> X, Vector<T> y) |
|
IEnumerable<DataSplitResult<T>> GetSplits(Matrix<T> X, Vector<T> y) |