Overview / Usage

In this project i have used two approaches to solve the problem.
The Problem:Sellers on amazon can describe one product in different ways so it is difficult to find out the product based on their names.
Solution:
This system calculate weights of the description of the product by doing sentiment analysis on the product description field and recommends the products which are most relevent products in terms of product description.
We can use the above system to help sellers to recommend more accurate products and increase their sales.

Methodology / Approach

1st approach : TF-IDF
In this approach i have calculated the term frequency of particular key words in each document and then calculated the weights using cosine similarity functions to determine the relevance of the products.
and then recommended 2 most relevant products.

2nd approach : Word2Vec
In this approach i used a pre-trained model of google news for the sentment analysis of the product description field.
Then i generated vector values for each product in the dataset and then compared it with the selected product to find the most near products in terms of weights. That will give us the most relevant products.
And then i recommended the top 2 products in the relevant products list.

Technologies Used

I used following technologies:

1. Machine Learning.
   2.Python 3
   3.Anaconda server
   4.Google News Vectors(Pre-trained ML model)

Repository

https://github.com/Nilex-Shinde/Recommender-system-using-ML