Dynamic Portfolio Analyst
Exploring the combination of time-series forecasting and financial news sentiment analysis.
A research prototype combining LSTM-based time-series forecasting with FinBERT sentiment analysis of financial news. The project explores whether information extracted from financial, business and technology news can complement historical price data when analysing short-term market movements.
Approach
- LSTM-based modelling of historical stock prices.
- FinBERT for extracting sentiment from financial news.
- News collection and processing using web scraping.
- A Flask and React interface for presenting model outputs and analysis.
Built with
Built on a fork of an existing repository; my contribution covers the forecasting and sentiment pipeline described above.