Sentiment-Analysis-of-Reviews-using-Machine-Learning-algorithms-on-Textual-data

CSCI 59000 BIG DATA ANALYTICS PROJECT

Description

• Programmed a XML parser in python using xml.etree.ElementTree package
• Text data is Pre-processed by removing special characters
• Word embeddings of text are created using Word2Vec tool and tokenized.
• A deep learning model is created using TensorFlow framework by implementing Long-short term memory (LSTM) based Recurrent neural networks.
• Bigrams are created for the text after undergoing pre-processing, which includes removing stop words and stemming.
• Naïve-bayes classification model is built using Bigrams and nltk package.
• Performance analysis of both models is done by drawing ROC curves, by comparing accuracies, and Area Under Curve.

Requirements

Python packages: numpy, tensorflow, matplotlib, nltk, sklearn, itertools

Dataset

sample Amazon XML dataset

LSTM RNN tensorflow model

used tensorflow

Bigrams Naive-bayes model

used nltk

Results

Predictive Model	Accuracy
Naive Bayes Classification	68.5
RNN using LSTM	70.83
Naive Bayes Classification with Bigrams	74.4

Naive Bayes Classification with Bigrams showed higest acuracy using nltk