Bull Stock

A Stock Price & Market Trend Prediction Engine which combines the Power of Historical Data with Wisdom of Social Media Buzz!

Comment

Inspiration

As a society, most people often imagine stock markets as a gambling space. However, we believe that through scientific study and data analysis, investing is a boon. Through this project, we want to want to encourage financial literacy and independence among all age groups.

What it does

We predict the stock prices and market trends of the top 500 S&P listed companies based on their historical data and sentiment obtained from tweets and articles of Wall Street journal. We have used Machine Learning and Deep Learning.

How we built it

1) We will train a deep learning model & use sentiment analysis in order to predict the stock price of 500 S&P listed companies.

2) We plan will use Yahoo Finance & Market Watch for collecting historical data & also collect data by web scrapping of Wall Street Journal.

3) Through our approach we aim to draw correlations between stocks of various companies to improvise our predictive ability.

4) We also plan to evaluate the Risk vs Return ratio of each stock.

5) We also aim to compute the daily return average of each stock. Also, evaluate the annual financial report for each company using Sentiment Analysis.

6) Using sentiment analysis we will classify sentiments of investors into 6 categories including: Negative, Positive, Uncertainty, Litigious, Constraining, and Interesting.

Challenges we ran into

1) Limitation of data 2)Analysing the weight of sentiments in tweets and wall street posts 3)Since we did not have data for every company for every market hour, we had to interpolate the curve in some cases. So the curve is largely discrete in nature. However continuous nature of the curve has been depicted.

Accomplishments that we're proud of

1) We have predicted the stock prices of the top 500 companies which are S&P listed. 2) We also made a user-friendly web app so that people can visualize stock trends and processes better. 3) We also predicted the image of a company in the market analyzing the sentiment of the tweets extracted.

What we learned

1) Machine Learning for the data obtained from Yahoo Finance. 2) Sentiment Analysis on the articles of Wall Street Journal. 3) RNN (specifically LSTM) for analyzing time-series data. 4) NLP (bag of words model) for data processing. 5) We also learned to calculate Jaccard similarity from the bag of words matrix. To get better insights we will generate sentiments from tf-idf model and calculated cosine similarity. 6) We also learned about the world of finance, which included concepts like EPS, Adjusted Closing Price, NAV, Stock Split etc.

What's next for Bull Stock

We want to make a more personalized wrapper for a more user-friendly experience. A recommendation engine that can recommend which stocks to buy. We also propose to built a connector for a Demat account so that users can directly perform all transactions on our website. We can also provide an API for other developers to get and use our predictions for other company analyses.

Share this project:

Updates