The Problem (Determine Stock Trend and Design Promotion Strategy Accordingly):

There are many news articles which report each company and its associated products or services. Do they have any impact to stock price? If the stock price trend is predicted as decline at a particular company, do salesmen have a good strategy to promote its product or service?

The Solution (Text Rank Semantic Analysis and Deep Q-Network (DQN) Reinforcement Learning):

We prepare web scrappers for five financial news sources. They are Business Week, CNN Money, Investor Guide, Market Watch, and Yahoo Finance. Applying text rank semantic algorithm (modified from webpages rank algorithm) to calculate documents score, such score will be treated as a feature parameter in later machine learning model. The accuracy of the prediction model is more than 60% and the model has increased the accuracy by 30% more by comparing with news random labeling with only 33% of accuracy.

Because marketing changes very fast, we hook tweepy to get most-updated marketing information for different geographical regions. We apply DQN to help salesmen design promotion route if the stock performance is bad. The reason for us to apply DQN is that it can handle dynamic change of marketing information from Tweets and geographical scalability issues very well.

Project Website

For more details, check out the project's website here.

Important Figures

System Architecture for Predicting Stock Trend from Financial News with TextRank Algorithm

Stock Trend Prediction Accuracy Comparison by News Sources

System Architecture for Promotion Route Design from Tweepy Marketing Information

Marketing Information from Tweepy by area, i.e., RED regions represent bad response, GREEN region represents good response. Each area is 1.4 X 1.4 square mile.

Successful Rate Comparison by Cities