The research aimed at finding correlation between tweets and future stock prices.
Details and results in the Project Description.
Reproduction and improvement results from the related article.
Prepare new dataset with corrected preprocessing and labeling.
Draft of train and test RNN (LSTM / GRU) with Optuna tuning. Currently not finished yet.
Train and test CatBoost and Tranformer (Roberta) models.
Train and test Naive Bayes model. Creating ensemble of 3 models (Naive Bayes, CatBoost, Transformer).
Calculate business metrics (profit, Sharpe Ratio, etc.) for the specified data and model parameters. The notebook could be launched either manually or in the batch mode (nb320_batch_launch.sh).