dwker/ttools
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

class weights for multiclass for training

Browse Source
This commit is contained in:
David Brazda
2024-12-05 11:12:35 +01:00
parent 277de3f73e
commit 68ea25d963
2 changed files with 51 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
setup.py
View File

@ -2,7 +2,7 @@ from setuptools import setup, find_packages
setup(
name='ttools',
version='0.7.98',
version='0.7.99',
packages=find_packages(),
install_requires=[
# list your dependencies here

69
ttools/models.py
View File

@ -906,22 +906,42 @@ class LibraryTradingModel:
model = set_gpu_params(model)
# Handle class imbalance for binary classification
# Determine class weights dynamically
class_counts = np.bincount(y_fold_train) # Counts the occurrences of each class
class_weights = {i: sum(class_counts) / (len(class_counts) * count) for i, count in enumerate(class_counts)}
# Check number of classes
if self.config.n_classes == 2:
n_0 = sum(y_fold_train == 0)
n_1 = sum(y_fold_train == 1)
# For binary classification, use scale_pos_weight
n_0 = class_counts[0]
n_1 = class_counts[1]
scale_pos_weight = n_0 / n_1
model.set_params(scale_pos_weight=scale_pos_weight)
# Train with early stopping
model.fit(
X_fold_train,
y_fold_train,
eval_set=[(X_fold_val, y_fold_val)],
#early_stopping_rounds=50,
#verbose=2
)
else:
# For multiclass classification, use sample_weight
sample_weights = [class_weights[label] for label in y_fold_train]
print("Model Training...")
if self.config.n_classes == 2:
# Train with early stopping
model.fit(
X_fold_train,
y_fold_train,
eval_set=[(X_fold_val, y_fold_val)],
#early_stopping_rounds=50,
#verbose=2
)
else:
# Train with early stopping
model.fit(
X_fold_train,
y_fold_train,
eval_set=[(X_fold_val, y_fold_val)],
sample_weight=sample_weights
#early_stopping_rounds=50,
#verbose=2
)
# Calculate score
if self.config.model_type == 'classifier':
pred = model.predict(X_fold_val)
@ -1029,15 +1049,26 @@ class LibraryTradingModel:
model.set_params(verbosity=1)
#balance unbalanced classes, works for binary:logistics
# Determine class weights dynamically
class_counts = np.bincount(y_train) # Counts the occurrences of each class
class_weights = {i: sum(class_counts) / (len(class_counts) * count) for i, count in enumerate(class_counts)}
# Check number of classes
if self.config.n_classes == 2:
n_0 = sum(y_train == 0) # 900
n_1 = sum(y_train == 1) # 100
scale_pos_weight = n_0 / n_1 # 900/100 = 9
model.set_params(scale_pos_weight=scale_pos_weight)
# For binary classification, use scale_pos_weight
n_0 = class_counts[0]
n_1 = class_counts[1]
scale_pos_weight = n_0 / n_1
model.set_params(scale_pos_weight=scale_pos_weight)
else:
# For multiclass classification, use sample_weight
sample_weights = [class_weights[label] for label in y_train]
print("Model Training...")
model.fit(X_train_scaled, y_train)
if self.config.n_classes == 2:
model.fit(X_train_scaled, y_train)
else:
model.fit(X_train_scaled, y_train, sample_weight=sample_weights)
else:
print("Using PROVIDED MODEL and SCALER")
model = self.use_model