Functions of synthetic data generators.¶

Created on Fri Apr 10 17:17:41 2020

This script is for generating synthetic data. You can use multi-class data to generate balance dataset.

Abdullah BAS abdullah.bas@boun.edu.tr BME Bogazici University Istanbul / Uskudar @author: abas

class trklearn.ADASYN¶

Bases: object

ADASYN(target, verbose=True, B=1, K=15, threshold=0.7)¶

This class is implementation of ADASYN.

Parameters

• Xnp (np.array) – Input array must be numpy

• target (np.array) – response array

• verbose (bool, optional) – If zero will not output any verbose. Defaults to True.

• B (int, optional) – B is the balance ratio that you want to reach. Defaults to 1.

• K (int, optional) – Numbers of neighbours. Defaults to 15.

• threshold (float, optional) – Activation threshold. Above this balance ratio function will not work. Defaults to 0.7.

Returns

[description]

Return type

[type]

fit_resample(target, B=1, K=15, threshold=0.7)¶

fit_resample stands for outputting the generated data combined with the input data

Parameters

• Xnp (float) – Input data

• target (float,int) – Corresponding response.

• B (int, optional) – Balance ratio. It is the threshold for the generated data. Defaults to 1.

• K (int, optional) – K-neigbours. Defaults to 15.

• threshold (float, optional) – It is the threshold for imbalance ratio. Function runs only below this ratio. Defaults to 0.7.

Returns

Output xnp [float]: Output target

Return type

[float]

class trklearn.ASUWO¶

Bases: object

ASUWO(target, n, k, irt, de=- 1, normalization=1)¶

ASUWO is supporting multi-class synthetic data generation.

Parameters

• (type (target) – all): Input array must be numpy array

• (type – all): Corresponding response to Xnp

• n ([type]) –

• k (int) – Neighbours number

• irt (float) – Minimum imbalance ratio targeted

• knn ([type]) – [description]

• de (int, optional) – . Defaults to -1.

• normalization (bool, optional) – Switch for normalization. Defaults to 1.

trklearn.EuclidianDistance(data1, data2)¶

Euclidian Distance implementation

Parameters

• data1 (float) – data point 1

• data2 (float) – data point 2

Returns

distance between two data points

Return type

[float]

class trklearn.MinMaxNormalization(data, axis=0)¶

Bases: object

Min-Max Normalization. Was using in conjunction of ADASYN to test results

data: Data to be normalized axis: 0 is by columns, 1 is by rows

returns: Normalized data

class trklearn.SMOTE¶

Bases: object

SMOTE(target, N=- 1, threshold=0.7, verbose=True)¶

Implementation of SMOTE algorithm

Parameters

• Xnp (float) – Input data

• target (float) – Target/Response array

• N (int, optional) – Maximum class. If it is not known left it to default. Defaults to -1.

• threshold (float, optional) – [description]. Defaults to 0.7.

• verbose (bool, optional) – [description]. Defaults to True.

Returns

xiP2 only generated synthetic data

Return type

[float]

fit_resample(target, N=- 1, threshold=0.7, verbose=True)¶

fit_resample is the function that outputs the input data and generated data combined in one array.

Parameters

• Xnp (float) – Input data

• target (float) – Target/Response array

• N (int, optional) – Class that has the maximum elements. Defaults to -1.

• threshold (float, optional) – Activation threshold of imbalance ratio. Below this threshold function will run. Defaults to 0.7.

• verbose (bool, optional) – Defaults to True.

Returns

Xnp output [float]: target output

Return type

[float]