Sentiment-Based Recommendation System

Table of Contents

• Overview
• Motivation
• Technical Aspects
• Project Setup
• Technologies Used
• Future Scope

Overview

The project focuses on building a sentiment-based recommendation system which will recommend upto 5 products for an existing user. The dataset used here is inspired from https://www.kaggle.com/datasets/datafiniti/grammar-and-online-product-reviews. It consists of a subset of the original dataset and has 30000 reviewes for 200+products.

Motivation

The aim of this project is to build an end-to-end recommendation system for an e-commerce platform that will recommend products to users based on their purchase history and previous ratings and reviews. This system will automate the process of product evaluation and selection, making it easier and more efficient for users to find products they are likely to enjoy. The system will analyze previous ratings and reviews from other users to determine the most popular and highly rated products, and then recommend these products to new users based on their purchase history and other factors. The end result will be a more personalized and efficient shopping experience for the users of the e-commerce platform.

Technical Aspects

1. Data Preprocessing and EDA

• the data pre-processing and EDA is covered in notebooks/DataPreprocessing.ipynb

2. Experimentation of different Ml models for sentiment analysis
   • Experimentation of sentiment prediction for Bag of words and TF-IDF representation using the below models.
   • Naive Bayes
   • Logistic Regression
   • XGboost The model experimentation is covered in notebooks/Sentiment_prediction.ipynb
3. Exploring collaborative filtering algorithms for recommendation
   • User-User based collaborative filtering
   • Item -Item based Collaborative filtering Recommendation models are covered in notebooks/Recommendation_System.ipynb
4. Model deployment using flask framework

Project Setup

Installing the packages

For running the project, please install the necessary packages mentioned in the requirements.txt using the below command

'pip install -r requirements.txt'

Run the main application

To run the flask app please run app.py file.

Future Scope

• The dataset had a huge class imbalance. This can be handled by adding synthetic data for negative reviews.
• In the data pre-processing notebooks only BOW and TF_IDF representation was experimented. Word embedding models can be explored , as it focuses on understanding the similarity between words.
• The project only recommends products for a user trained in the dataset. Better algorithms and approaches can be used to build a more realistic recommendation system.