Record Class BSO

java.lang.Object

java.lang.Record

smile.classification.resampling.BSO

Record Components:

data - the augmented feature matrix (original + synthetic samples).

labels - the augmented labels (original + synthetic sample labels).

public record BSO(double[][] data, int[] labels)
extends Record

Borderline Shifting Oversampling (BSO).

BSO extends SMOTE by combining the borderline sample identification from Borderline-SMOTE (Han et al., 2005) with a shifting synthesis mechanism. Instead of interpolating between two minority samples, BSO shifts each borderline minority sample toward the local majority centroid, placing synthetic points directly in the critical region between the classes.

Borderline classification — for each minority sample xᵢ, its m nearest neighbors in the entire dataset are found and the number of majority neighbors mʼ is counted:

• mʼ = m → NOISE: entirely surrounded by majority; never used as a seed.
• m/2 ≤ mʼ < m → DANGER: on the class boundary; the primary seed candidates for synthesis in both DANGER and DANGER_AND_SAFE.
• mʼ < m/2 → SAFE: deep in the minority region; skipped in DANGER, used as secondary seeds at a reduced shift rate in DANGER_AND_SAFE.

Shifting mechanism — for a seed xᵢ, its kMaj nearest majority neighbors are found and their centroid cᵢ is computed. A synthetic sample is generated as:

  x_syn = xᵢ + α · (cᵢ − xᵢ) + jitter

where α ∈ (0, 1) is the shift factor and jitter is a small per-dimension noise scaled to a fraction of the shift magnitude to prevent all samples from collapsing to the same point. SAFE seeds in DANGER_AND_SAFE use α/2 to shift more conservatively.

When kMaj = 0, the shifting mechanism is disabled and the algorithm falls back to standard Borderline-SMOTE interpolation: a synthetic sample is placed at a random position on the line segment between the seed and one of its randomly selected minority nearest neighbors.

If no DANGER samples exist, all minority samples are used as seeds (fallback identical to plain SMOTE but with the selected synthesis strategy).

Index selection
When the dimensionality d ≤ highDimThreshold (default 20) a KDTree is used; otherwise a RandomProjectionForest is used.

Limitations

• Feature spaces must be entirely continuous (no categorical features).
• A large α may push synthetic samples deep into the majority region, potentially introducing ambiguous training examples.

References

1. Malhat M.G., Elsobky A.M., Keshk A.E. et al. An approach for handling imbalanced datasets using borderline shifting. Sci Rep (2026).
2. N. V. Chawla, K. W. Bowyer, L. O. Hall and W. P. Kegelmeyer.
3. H. Han, W. Wang and B. Mao. Borderline-SMOTE: A New Over-Sampling Method in Imbalanced Data Sets Learning. ICIC, LNCS 3644, 878–887, 2005.
4. N. V. Chawla, K. W. Bowyer, L. O. Hall and W. P. Kegelmeyer. SMOTE: Synthetic Minority Over-sampling Technique. JAIR 16:321–357, 2002.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final record	BSO.Options	BSO hyperparameters.
static enum	BSO.Strategy	BSO strategy controlling which minority samples are used as seeds.

Constructor Summary

Constructors

Constructor	Description
BSO(double[][] data, int[] labels)	Creates an instance of a BSO record class.

Method Summary

Modifier and Type	Method	Description
double[][]	data()	Returns the value of the data record component.
final boolean	equals(Object o)	Indicates whether some other object is "equal to" this one.
static BSO	fit(double[][] data, int[] labels)	Applies BSO to the given dataset with default BSO.Options.
static BSO	fit(double[][] data, int[] labels, BSO.Options options)	Applies BSO to the given dataset.
final int	hashCode()	Returns a hash code value for this object.
int[]	labels()	Returns the value of the labels record component.
int	size()	Returns the total number of samples after resampling.
final String	toString()	Returns a string representation of this record class.

Methods inherited from class Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Constructor Details

BSO

public BSO(double[][] data,
 int[] labels)

Creates an instance of a BSO record class.

Parameters:

data - the value for the data record component

labels - the value for the labels record component

Method Details

size

public int size()

Returns the total number of samples after resampling.

Returns:

the number of rows in data.

fit

public static BSO fit(double[][] data,
 int[] labels)

Applies BSO to the given dataset with default BSO.Options.

Parameters:

data - the input feature matrix; each row is an observation.

labels - the class labels corresponding to each row of data.

Returns:

a BSO instance holding the augmented data and labels.

fit

public static BSO fit(double[][] data,
 int[] labels,
 BSO.Options options)

Applies BSO to the given dataset.

The minority class (label with the fewest occurrences) is identified automatically. Each minority sample is classified as NOISE, DANGER, or SAFE based on its neighborhood composition. Synthetic samples are then generated by shifting DANGER (and optionally SAFE) samples toward the local majority centroid.

Parameters:

data - the input feature matrix; each row is an observation.

labels - the class labels corresponding to each row of data.

options - the hyperparameters.

Returns:

a BSO instance holding the augmented data and labels.

Throws:

IllegalArgumentException - if data and labels have different lengths, if the input is empty, or if the minority class has fewer samples than options.m() + 1.

toString

public final String toString()

Returns a string representation of this record class. The representation contains the name of the class, followed by the name and value of each of the record components.

Specified by:

toString in class Record

Returns:

a string representation of this object

hashCode

public final int hashCode()

Returns a hash code value for this object. The value is derived from the hash code of each of the record components.

Specified by:

hashCode in class Record

Returns:

a hash code value for this object

equals

public final boolean equals(Object o)

Indicates whether some other object is "equal to" this one. The objects are equal if the other object is of the same class and if all the record components are equal. All components in this record class are compared with Objects::equals(Object,Object).

Specified by:

equals in class Record

Parameters:

o - the object with which to compare

Returns:

true if this object is the same as the o argument; false otherwise.

data

public double[][] data()

Returns the value of the data record component.

Returns:

the value of the data record component

labels

public int[] labels()

Returns the value of the labels record component.

Returns:

the value of the labels record component