README.md

Financial-Document-Sentiment-Analysis

Machine Learning Techniques for Financial Sentiment Analysis

The modern stock market generates large amounts of numerical data regarding specific stocks. Along with this data, the opinion of various experts/investors are available freely on the internet. Quantifying the general sentiment regarding the stock based on such data will help in predicting the trend (Bearish/Bullish) of the stock price. This can be achieved with the help of several supervised machine learning techniques trained using a corpus of positive as well as negative financial news/blogs/reviews. Some of the major challenges are identifying the appropriate sources for stock related data, pre-processing unstructured text and converting the same to structured data in a scalable and efficient manner.

Work done until now:
  1. Scrapped Moneycontrol website and created dataset for financial data [Web Scrapper MoneyControl.ipynb]
    1.1. [MoneyControl-First-1-20-Pages.csv] was created by the above notebook
    1.2. [Temporary File] A new set of articles was fetched using same scraper not yet attached to main data
  2. Applied text classification model Support Vector Classifier for Sentiment Analysis [Text Classification Money Control.ipynb]
  3. Applied Sequence Classification with LSTM Model [LSTM Sentiment Analysis.ipynb]
    3.1. Model Generated [money_control_model.h5]
  4. Vader Sentiment Analysis