Application of both Machine Learning (ML) and sentiment analysis from
microblogging services has become a common approach for stock market prediction.
In this thesis, Microsoft’s stock movements are analyzed using historical and sentiment
data. In particular, 90.000 tweets were collected from Twitter and 7.440 tweets from
StockTwits covering the period from 16 – 07 – 2020 to 31 – 10 – 2020. Historical data
were also mined from the Finance Yahoo website at the same period. The sentiment
analysis of social media data was conducted using two Python libraries including
TextBlob and VADER (Valence Aware Dictionary and sEntiment Reasoner). We also
implemented multiple machine learning models including KNN, SVM, Logistic
Regression, Naïve Bayes, Decision Tree, Random Forest and MLP. Our results indicate
that when using tweets from Twitter with VADER as sentiment analysis tool, SVM is
the ML algorithm which gives the highest f-score equal to 75.9% and Area Under Curve
(AUC) equal to 65%.
Collections
Show Collections