Predict stock sentiment from Social Media
Using labelled data from StockTwits (tweet-style social media site discussing stocks), trained an LSTM recurrent neural network on the labelled data.
Trained model has a ±70% accuracy. Example output for the tweet: "Google is working on self driving cars, I'm bullish on $goog"
Rating | Probability |
---|---|
0 (negative) | 0% |
1 | 2% |
2 | 1% |
3 | 71% |
4 (positive) | 26% |