Optimizing Digital Advertising with Deep Learning: Classifying News Articles
- Optimized a deep learning DistilBERT-based model for health and wellness relevance prediction, achieving 95% accuracy by fine-tuning hyperparameters like learning rate and batch size with the one-cycle policy.
- Bag-of-Words Approach: Applied TF-IDF vectorization for feature extraction and tested multiple classifiers (Ridge, Passive-Aggressive, Random Forest).
- Accuracy with SGD Classifier: Achieved 89% accuracy using Stochastic Gradient Descent (SGD) classifier with the Bag-of-Words approach.
- Performance Improvement: The deep learning model outperformed the traditional approach, improving accuracy by 6% (from 89% to 95%).
- Decision-Making: Experimented with various classifiers to compare results, ultimately selecting the deep learning model for its superior performance.
