This is a preliminary work in progress repo for
[1] experiments in stacking LASSO + XGBoost + text2vec in R
[2] prediction of Amazon review sentiments
The method landed me 1st place in a Kaggle In Class competition:
https://inclass.kaggle.com/c/irgn452-text-mining-task
Training data is Amazon product reviews with either a positive or a negative sentiment, hand-coded.
Goal is to assign sentiment scores to unmarked test data.
Value: what can Amazon vendors can improve the quality of their services and products if then can automatically scores review sentiments.
Inspired by an assignment in Big Data Analytics class at UC San Diego.