Sentiment analysis on hackerearth predict happiness challenge
accuracy - 88.592% Model - Ensemble of Random Forest, Gradient Boost, Logistic Regression and Naive Bayes
Real Time Twitter Sentiment Analysis
So I pickled the models created by hackerearth competition and used it for making predictions in another cool thing - Twitter Sentiment Analysis but so as to make it look professional & meaningful I made it end to end.
TWITTER ---tweets--stream----> PYTHON MACHINE LEARNING CODE---sentiment--tagged--tweets---> mySQL DB
Hence as many times I will run it will keep accumulating results and simulataneously store in database can serve for future modelling as well (I can manually tag the incorrect predictions and make model stronger) . It can stream live tweets thanks to awesome twitter api and predict sentiment of it! I had pretty good results on #flipkart #bigbillionday sale vs #amazon #greatindainsale which analysed twitter sentiments of customers.