search

TSL-123 / SentimentDrivenStrategy

0 stars 0 forks source link

Final Peer Review (wl596) #18

Open WanlinLi opened 6 years ago

WanlinLi commented 6 years ago

This project aims to build model to analyze the sentiment of news automatically. The data comes from a variety of sources including Thompson Reuters historical real-time news, WRDS TAQ Database, WRQS CRSP Database, and Fama French Database.

Things that I like:

  1. The team made a very clear explanation about their goal and the worth of this project.
  2. The group gave a clear explanation about the CAPM model, which they used to compute abnormal return. And they used rolling regression to estimate beta as a time varying variable.
  3. The report includes a detail about how the team gradually improved their trading strategy from a negative growth to a large positive growth, which is quite impressive.

Things that may need improvement:

  1. The 768% (or 500%) growth in 8 months seems too good to be true. The group may add some more verifications or explanations about this result, in order to make people believe in it.
  2. Some data visualization techniques may be helpful for people to get a better understanding about the dataset.
  3. The team only use a few techniques from what we learnt from class.