Read Twitter json file from database every hour
Clean and tokenized the tweets
Using trained model to classify Tweets into relevant and irrelevant
Saving Tweets into dataframe with specific confidence level (>0.99 | <= 0.01) for re-training purpose
Doing Sentiment Analysis on relevant Tweets
Sleep for an hour then re-do the whole process above
Re-train model every Sunday 10am