Sentiment analysis is performed on amazon review data This code is for extracting sentiment from amazon review data set provided by UCSD.
Dataset used - Amazon review data - UCSD dataset
Dataset Citations
Ups and downs: Modeling the visual evolution of fashion trends with one-class collaborative filtering R. He, J. McAuley WWW, 2016 pdf
PROCEDURE FOLLOWED
-
Sentiment is not directly provided in the data set, so polarity is obtained for each review using text blob library
-
Dataset is divided into test and train
-
Following are the classifiers used for the analysis purpose
- Naive Bayes
- Decision Tree
- Adaboost Classifier
- Random Classifier
- Multi Layer Perceptron
RESULT MLP outperfomed other classifiers