harshhmaniya/fake-news-classification-distilbert-fine-tuned

Fake News Classification DistilBERT Fine-Tuned

This model is a fine-tuned version of distilbert-base-uncased on an unknown dataset. It achieves the following results on the evaluation set:

Model Architecture

• Base Model: TFDistilBertForSequenceClassification
• Number of Labels: 2
• Framework: TensorFlow with Hugging Face Transformers

Training Configuration

• Optimizer: Adam
• Learning Rate: 0.001
• Loss Function: Sparse Categorical Crossentropy
• Epochs: 3
• Batch Size: 8
• Data Split:
  • The dataset was loaded from the Kaggle fake news detection files (“True.csv” and “Fake.csv”).
  • (Note: The notebook internally handles the train/validation/test split.)

Performance Metrics

After fine-tuning, the model achieved the following evaluation results (as logged during training):

• Training Accuracy: ~99.5%
• Validation Accuracy: ~99.8%
• Testing Accuracy: ~99.7%

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 0.001, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
• training_precision: float32

Framework versions

• Transformers 4.48.3
• TensorFlow 2.18.0
• Tokenizers 0.21.0