Well done, the website looks solidly made. This is my first time seeing sentiment illustrated with candlesticks, I find that very interesting.
My first instinct is I want to see index ticker symbols overlaid on top of the sentiment chart to how well they do or don't correlate. Just showing periods of on/off correlation between tickers and sentiment in general would be an interesting visualization. Have you considered this?
Thank you so much! This was actually my first time working with react, or really any front end design in general. Spent a lot of time on it.
The candlesticks do actually track the price data just like most stock charts. The purple line does however track the rolling sentiment, and the 'volume' bars represent the volume of bullish / bearish trades actually mentioned
A lot of people do this but the major bench mark everyone compares against it the S&P 500.
My recommendation is take your data, find 10-years if you can, and then rank the top 100, or 500 stocks, I recommend 100. Now create a portfolio which rebalances itself monthly based on the sentiment from each network.
So you can have the WSB69, TWTR140, ST50, and such, to see which one is the most accurate. From my testing with sentiment analysis ages ago, I believe it doesn't work, but there are people who would jump onto those because it's fun for them.
This is pretty cool. I am also learning react (just started). Did you use something pre-built for the options table?
As for NLP, can models understand blocks of text?
I have t kept up with progress of NLP but I thought sentiment analysis was just based on keywords and labeled data.
So far it’s seems the most bullish sentiments are predicting the market incorrectly. I guess that’s why they say go against the popular sentiment on wallstreet bets.
I used a package called data-ui for the distributions and histograms, would recommend.
The NLP I use is all custom and does a kind of hierarchical pattern matching. It identifies low level meaing and uses that as building blocks for higher level.
And yes it has been my observation that the market often reverses for the most popular plays
[+] [-] KukiAirani|5 years ago|reply
My first instinct is I want to see index ticker symbols overlaid on top of the sentiment chart to how well they do or don't correlate. Just showing periods of on/off correlation between tickers and sentiment in general would be an interesting visualization. Have you considered this?
[+] [-] electricwallaby|5 years ago|reply
The candlesticks do actually track the price data just like most stock charts. The purple line does however track the rolling sentiment, and the 'volume' bars represent the volume of bullish / bearish trades actually mentioned
[+] [-] shyn3|5 years ago|reply
My recommendation is take your data, find 10-years if you can, and then rank the top 100, or 500 stocks, I recommend 100. Now create a portfolio which rebalances itself monthly based on the sentiment from each network.
So you can have the WSB69, TWTR140, ST50, and such, to see which one is the most accurate. From my testing with sentiment analysis ages ago, I believe it doesn't work, but there are people who would jump onto those because it's fun for them.
[+] [-] watcher2020|5 years ago|reply
So far it’s seems the most bullish sentiments are predicting the market incorrectly. I guess that’s why they say go against the popular sentiment on wallstreet bets.
[+] [-] electricwallaby|5 years ago|reply
The NLP I use is all custom and does a kind of hierarchical pattern matching. It identifies low level meaing and uses that as building blocks for higher level.
And yes it has been my observation that the market often reverses for the most popular plays
[+] [-] loremipsium|5 years ago|reply