Code implementation of the sentiment score extraction models can be found in another repository

First, it is needed to clean the raw text data and then generate useful features which will be exploited to the model (Stocktwits_Data_Preparation.py). Then, the missing value, constant, and correlated feature analysis have to be done (Microblog_Text_Data/Stocktwits_Data_Cleaning.ipynb). Finally, the prepared data is exploited as the input to the trained ensemble-model and then the extracted sentiment score is appended to the historical stock data (Stocktwits_Sentiment_Prediction_Join_Data.py).