stockwits-anayltics

One million StockTwits messages collected via API and stored in MongoDB, preprocessed with emoji/punctuation preservation, and classified using TF-IDF with Logistic Regression, SVM, Naïve Bayes, Random Forest, and MLP; extended with n-gram analysis, forecasting regressions, and event-study tests linking sentiment to short-term stock returns