Business Intelligence Visualization using Deep Learning Based Sentiment Analysis on Amazon Review Data

This repository contains the code for the paper titled "Business Intelligence Visualization using Deep Learning Based Sentiment Analysis on Amazon Review Data", which has been published in the proceedings of the 12th IEEE International Conference on Computing, Communication and Networking Technologies (ICCCNT) held at IIT Kharagpur. (Click here to read the research paper)

This paper proposes the use of machine learning and deep learning algorithms to classify the sentiment of 60,000 customer reviews on Amazon products as positive or negative. The dataset named ‘Amazon Reviews: Unlocked Mobile Phones’ was provided by PromptCloud in Kaggle. The paper also uses business intelligence to effectively visualize the results in the form of bars, graphs and charts. Sentiment analysis was successfully performed with a highest accuracy of 98.51% obtained from the BERT deep learning model.

The following machine learning algorithms were used for performing sentiment analysis:

• Decision Tree
• Logistic Regression
• Stochastic Gradient Descent
• Multinomial Naive Bayes
• Support Vector Machine

The following deep learning algorithms were used for performing sentiment analysis:

• Bidirectional Encoder Representations from Transformers (BERT)
• Long Short Term Memory (LSTM)

PowerBI from Microsoft was used as the business intelligence tool.

Authors: Harish Balasubramaniam, Zeel Desai, Karishma Anklesaria