-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
95a6961
commit 4e18fda
Showing
3 changed files
with
70 additions
and
4 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,5 +1,5 @@ | ||
.venv/ | ||
__pycache__/ | ||
**/__pycache__/ | ||
.vscode | ||
build/ | ||
dist/ |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,66 @@ | ||
import numpy as np | ||
from math import log | ||
from itertools import compress | ||
|
||
from cobraclassifier import classifier_cobra as cobra | ||
from cobraclassifier import edited_knn, near_miss_v1, near_miss_v2, near_miss_v3, tomek_link, condensed_knn, knn_und | ||
|
||
|
||
class CobraBoost: | ||
def __init__(self, X, y, machines, undersampling_method): | ||
self.X = X | ||
self.y = y | ||
|
||
self.model = cobra(machines = machines) | ||
self.majority_class_label = int(sum(y) > 0.5 * len(y)) | ||
self.undersampling_method = undersampling_method | ||
|
||
self.weight_update = 0 | ||
self.init_w = 1.0 / len(self.X) | ||
self.weight = np.full(len(self.X), self.init_w) | ||
|
||
|
||
def learn_parameters(self, iterations): | ||
verdict = self.undersampling_method.undersample(self.X, self.y, self.majority_class_label) | ||
X_undersampled, y_undersampled = self.X[verdict, :], self.y[verdict] | ||
|
||
for t in range(iterations): | ||
print("[Testing]: Executing the iteration - {} of CobraBoost".format(t + 1)) | ||
|
||
self.model.fit(X_undersampled, y_undersampled, sample_weight = self.weight[verdict]) | ||
|
||
flag = self.y != self.model.predict(self.X) | ||
loss = sum(list(compress(self.weight, flag))) | ||
|
||
alpha = loss / (1 - loss) | ||
|
||
if alpha <= 0: | ||
alpha = 0.0000001 | ||
else: | ||
try: | ||
alpha_hat = 0.5 * (np.log(1 - loss) - np.log(loss)) | ||
except: | ||
alpha_hat = 0 | ||
|
||
self.weight = self.weight * np.exp(-alpha_hat * self.y * self.model.predict(self.X)) | ||
self.weight = self.weight / self.weight.sum() | ||
|
||
self.weight_update = alpha | ||
|
||
|
||
def predict(self, test_data): | ||
n = len(test_data) | ||
predicted_labels = np.zeros(n) | ||
|
||
for i in range(n): | ||
positive_score, negative_score = 0, 0 | ||
|
||
if self.model.predict(test_data[i].reshape(1, -1)) == 1: | ||
positive_score += log(1/self.weight_update) | ||
else: | ||
negative_score += log(1/self.weight_update) | ||
|
||
if negative_score <= positive_score: | ||
predicted_labels[i] = 1 | ||
|
||
return predicted_labels |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -2,13 +2,13 @@ | |
setup( | ||
name = 'cobraclassifier', | ||
packages = ['cobraclassifier'], | ||
version = '1.3', | ||
license='MIT', | ||
version = '1.4', | ||
license = 'MIT', | ||
description = 'COBRA for classification tasks (on Imbalanced Data)', | ||
author = ['Dr. Arabin Kumar Dey', 'Vishisht Priyadarshi', 'Aadi Gupta', 'Tejus Singla', 'Shashank Goyal'], | ||
author_email = '[email protected]', | ||
url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6', | ||
download_url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6/archive/refs/tags/v1.1.tar.gz', | ||
download_url = 'https://github.com/vishishtpriyadarshi/MA691-COBRA-6/archive/refs/tags/v1.4.tar.gz', | ||
keywords = ['Classification', 'Imbalanced Data', 'Machine Learning'], | ||
install_requires=[ | ||
'numpy', | ||
|