This project aims to investigate the prediction power of public sentiment on FANG, i.e Facebook, Amazon, Netflix and Google, extracted from Economics Subreddit and New York times by comparing the prediction accuracy of different machine learning methods.
We conclude that simpler models, such as linear regression and random forest better predict (63.16% accuracy) on stock return's growth trend using sentiment, largely due to high correlations between sentiment effects and stock prices.
RESULT:
ASSUMPTION:
DATA:
MODEL:
- Abstract
- Data Source
- Exploratory Data Analysis
- Benchmark
- Sentiment Analysis
- Trading Strategy
- Conclusion
- Team
Sorted New York Times Data
Stock Price from Nov 26th 2018 to Nov 11th 2019:
Emotional Analysis with NRC emotion lexicon: counting 8 emotions and 2 sentiments
def return_emotions(stock_df):
count = 0
columns = ['anticipation', 'sadness', 'joy', 'negative', 'trust', 'positive', 'surprise', 'disgust', 'anger', 'fear']
emotions = pd.concat([pd.DataFrame(emotion_analyzer(i), columns=columns)
for i in stock_df['lead_paragraph']], ignore_index=True)
return emotionsNew York Times Headline Daily Sentiment Daily
Daily Sentiemnt
| Danielle (zw2631) | Mary Shao |
|---|---|
 |
 |