Financial News Sentiment Analysis

A Python-based project that analyzes financial news headlines and predicts the overall sentiment for companies using NLP models.

Live Demo

Try it out:

pip install -r requirements.txt
python run_analysis.py

Motivation

In today’s fast-paced world, staying updated on financial news is crucial. Stock prices are heavily influenced by company headlines, but manually tracking this information is challenging

Years ago, I had the idea to use models to predict stock prices, but it didn’t work. I realized that a key factor influencing stock prices is the news headlines of the company.

That’s exactly why this project came to be – to give a clear picture of market sentiment by analyzing financial news headlines.

You cannot predict the future. -Stephen Hawking (A Brief History of Time)

Features

• Automatically scrapes financial news articles.
• Stores articles and sentiment in a PostgreSQL database
• Use a pretrained model to classify headlines as positive, negative or neutral
• Use a custom model to do Named-entity recognition (NER) on companies

Structure

• scraper.py: Scrapes news articles, inserts them into a PostgreSQL database, and triggers the Kaggle notebook.
• financial-news-analyzer.ipynb: Runs the models on Kaggle and inserts the results into the same PostgreSQL database.
• kaggle.ipynb: File used to fine-tune the model
• api/index.py: Flask web application that serves the demo website and exposes API endpoints for retrieving articles and sentiment summaries from the database
• public/index.html: HTML page for the demo site.

Models

• Based on DistilBERT, fine-tuned for financial news sentiment classification.
• You can find the sentiment model here