100 Days of ML Days 138-145 : JanataHack NLP Hackathon by Analytics-Vidhya

100 Days of ML - Days 138 - 145

JanataHack NLP Hackathon by Analytics-Vidhya

Participating and working on the Steam reviews Sentiment Analysis problem.
Task: Predicting from text reviews whether a user recommends the game on Steam.

 Approaches:

1) Used word count(most frequent words in the two categories) and symbol count as features and trained an XGBoost model (0.82 F1-score)
https://github.com/hithesh111/HithBeyond100/blob/master/Hackathons/JanataHackNLP/approach1.ipynb

2) Created word count(most frequent words in the two categories) and trained a Naive Bayes Model (0.79 F1-score)
https://github.com/hithesh111/HithBeyond100/blob/master/Hackathons/JanataHackNLP/approach2.ipynb

3) Used tf-idf vectorizer to create features and then trained a Linear SVM model (0.84 F1-score)
https://github.com/hithesh111/HithBeyond100/blob/master/Hackathons/JanataHackNLP/approach3.ipynb