This project aims to do Sentiment Analysis on News Headlines by doing a short Exploratory Data Analysis then creating Classification Models. Two methods for vectorization were used, TF-IDF and Sentence Transformer embeddings. In addition, the classification models that were explored are SVM, Random Forest and LightGBM.
To fully understand how the selected model classifies the news headlines, model explainability using LIME was done. SHAP was also considered but running the explainability took hours.
[ ] Use transformer model as classification model
[ ] Identify news headline topics using topic modeling