Financial News Headlines Sentiment Analysis

Collect financial news headlines by scraping seekingalpha.com
Perform sentiment analysis on headline news using NLP models
Train models on larger dataset to improve accuracy

Prerequisites

Python 3.8.6
Python libraries pip install -r requirements.txt
Google Chrome driver

Data

Facebook's historical price data From Yahoo Finance
Web scraper for headline news
Processed news headlines - data
1 million finance news headlines From Kaggle
Processed news headlines (large)

Models

BiLSTM (trained on large dataset) - 60% accuracy
BERT - 52% accuracy
BiLSTM model overfit

Results

Saved model
Predicted results

Acknowledgement

Updated adam0ling's BERT model to work with the lastest version

License