Through this project we pushed sentiment analysis predication to predict the actual class level rating [1-5] rather than other emotion predication models. Below are other implementations executed on this data set:
This technique constructs classifiers: the model assigns class labels to problem instances. Naive Bayes Classifiers can be trained very efficiently in a supervised learning setting. It is a probabilistic classifier, and calculates the probability that a sentence exists, such as "Loved the product!" An advantage of naive Bayes is it only requires a small number of training data to estimate the parameters necessary for classification.
Khasim Pannegatla has an example of using this with the Amazon Review text on Kaggle, you can check it out through the link below.
A 'specialized' version of Naive Bayes, MultinominalNB is designed more for text documents. MutlinominalNB explicitly models the word counts and adjusts the underlying calculations. It estimates the conditional probability of a particular word given in a class as the relative frequency.
Yashraj Nigam predicts sentiment of 0,1, and 2 rather than Ordinal Classification using Multinomial NB, you can check out his results through the link below.
The below are successful model examples utilizing the Amazon data set by reducing predication to 'positive' or 'negative' (label_1 and label_2).
__label__1 - groups 1-star and 2-star ratings
__label__2 - groups 4-star and 5-star ratings
**it is assumed that he has removed 3-star ratings from the model to assume most 3-star ratings are neutral in sentiment.
SVM is a supervised (feed me) type of machine learning that can be used for both classification or regression challenges.` SVM performs classification by finding the hyper-plane that differentiates the classes plotted in n-dimensional space.