How xraptorgg/finbert-lstm works
Project type and purpose
• This is a data‐science pipeline for stock price prediction based on news sentiment.
• It implements steps from news and stock data collection, through cleaning and sentiment analysis (using FinBERT), to model training (MLP, LSTM, LSTM+BERT) and final analysis.
File structure
• Flat layout with numbered Python scripts indicating sequential stages:
1_news_collection.py
2_stock_data_collection.py
3_news_data_cleaning.py
4_news_sentiment_analysis.py
5_MLP_model.py
6_LSTM_model.py
7_lstm_model_bert.py
analysis.py
• Data artifacts (CSV files) store intermediate outputs: news.csv, news_data.csv, sentiment.csv, stock_price.csv
• No framework subfolders—this is a linear pipeline, not an MVC or microservice setup.
README for additional insights
• Purpose: replicate the paper “Deep Learning based stock price prediction using News Sentiment Analysis.”
• Key technologies: Python, FinBERT for sentiment, deep neural networks (MLP, LSTM variants).
Detailed guidance
• If desired, note compute resources (GPU, CPU) under model components.
• Stage’s input and output artifacts are shown explicitly.
• Keep the flow left‐to‐right (ingest → transform → train → evaluate).