I’ve built a (non-monetized, infrequently updated, portfolio project) web application that tracks the pulse of the Bitcoin community. The application, which I’ve named TokenSense, leverages a variety of price information, activity metrics, and sentiment metrics to provide actionable insights into the general sentiment of buyers and sellers of Bitcoin. I have also employed clustering techniques to group historical dates by their different levels of sentiment and activity. Has the market been overheated lately? Is it flatlining? TokenSense can provide a clearer picture, and it can also go deeper. Using topic modeling and sentiment analysis, TokenSense calculates and visualizes the prevalance and sentiment of news article topics across time.
Read. Watch. Review.
-
Watch the powerpoint presentation.
-
Read the full project write-up on Medium.
-
Watch the app demo and code review.
-
Review the project code on my Github.
-
![]()
Take a deep dive into a topic, like Institutional Investing.
-
![]()
The Bull Run was the big topic, until it wasn't.
-
![]()
From collection to production.
-
![]()
Making sense of 20,000 news articles.
-
![]()
Principal Component Analysis on dates clustered by sentiment.
-
![]()
LDA, maybe TMI.
