Financial Sentiment Analysis with Traditional & LLM Models

This project evaluates multiple approaches to sentiment classification on financial text data — ranging from traditional ML models to zero-shot large language models (LLMs).

---

Overview

We benchmark the performance of several models on financial sentiment classification, comparing accuracy, F1 scores, and per-source breakdowns across diverse datasets.

Models Compared

Model	Type	Notes
Logistic Regression	Traditional ML	TF-IDF + LogisticRegression
FinBERT	Transformer	Finetuned BERT on finance texts
SpaCy + TextBlob	Rule-based	Polarity → Label mapping
o4-mini (Azure)	LLM	o4-mini via Azure OpenAI

---

Folder Structure

project-root/
├── notebooks/               ← Main experiment notebooks
│   ├── 1_data_preprocessing.ipynb
│   ├── 2_logistic_regression_baseline.ipynb
│   ├── 3_finbert_inference.ipynb
│   ├── 4_spacy_textblob.ipynb
│   ├── 5_o4mini_inference.ipynb
│   └── demo.ipynb          ← 📊 Summary notebook with visuals
│
├── scripts/                ← Reusable utilities
│   ├── metrics.py
│   ├── preprocessing.py
│   └── plot_utils.py
│
├── data/
│   └── processed/          ← Cleaned CSVs from preprocessing
│
├── models/                 ← Saved predictions & classification reports
│
└── README.md

---

📦 Setup

git clone https://github.com/your-username/financial-sentiment-analysis.git
cd financial-sentiment-analysis
pip install -r requirements.txt

Requires: Python 3.10+, Jupyter, scikit-learn, transformers, spacy, seaborn

Optional:

python -m spacy download en_core_web_sm

---

Run Notebooks

Start by running the following notebooks in order:

1. 1_data_preprocessing.ipynb – Loads and prepares datasets
2. 2_logistic_regression_baseline.ipynb
3. 3_finbert_inference.ipynb
4. 4_spacy_textblob.ipynb
5. 5_o4mini_inference.ipynb
6. demo.ipynb – 📊 Visit this first! Summary visualizations and model comparison.

---

📊 Demo Preview

A side-by-side comparison of model F1 scores by source and sentiment class is provided in demo.ipynb.

---

Notes on Azure OpenAI

To run 5_o4mini_inference.ipynb, you need:

• Azure OpenAI access
• A deployed o4-mini model (deployment name: o4-mini)
• Your endpoint and key in the notebook (suggest using environment variables)

---

📜 License

This project is for educational and research purposes only. You may adapt and share with attribution.

---

Acknowledgements

• FiQa 2018
• FinBERT
• Financial PhraseBank
• Azure OpenAI