A flask based implementation for stock sentiment analysis. Uses randomforest with count vectorizer to get the sentiment. Results are given below.
Dataset Consists of 3 columns
Date, Label, Headline. Further analysis can be found in StockSentimentAnalysis.ipynb
| Classifier | Accuracy Avg(Precision) |
|---|---|
| Random Forest with countVectorizer | 86.0% |
| Random Forest with TfidfVentorizer | 83.8% |
| Naive Bayes MultinomialNB | 85.1% |
| XGBClassifier (check *.ipynb) | 84.9% |
| Logistic Regression | 85.4% |
| LinearSVC | 84.3% |
| SGDClassifier | 84.3% |
- Try out LSTM and other methods.
- Deploy it.
(4/07/2020) Switched to nltk and VADERsentiment (Valence Aware Dictionary and sEntiment Reasoner). This renders the mainFile useless but the data-analysis is available at StockSentimentAnalysis.ipynb
Credit goes to krishnaik06 who's livestream got me to do this. You can check his work here