Overview

One of client’s eCommerce website provides a wide range of product catalog that keeps growing gradually following the latest trends and requirements of new edge customers. To better understand and serve the customers’ we implemented a product category classification feature, using machine learning algorithms. These machine learning algorithms make use of a text description of the products in order to predict their categories automatically.

The Requirement

  • The requirement was to predict the category of a product based on its image using one or several images associated with the product along with its product id where the model can predict the correct product
  • Build a system that can forecast a standardized category
  • Create a classifier based on supervised learning

The Challenges

  • The absence of complete product information and unstructured titles of products
  • Lowercasing, eliminating special characters, remove words like ‘and,’ ‘by,’ ‘for,’ etc
  • Lack of consistency in product titles throughout the website

The Solution

  • We offered the solution using Data science - NLP to automatically choose the best category for each new product and improve the buyer’s shopping experience.
  • We applied machine learning algorithms - Text-based Classification to the text description of the products to automatically predict their category.

Product Categorization Engine

acr

The Development

Training Phase

  • We used dense layer,Recurrent Layers,Convolutional Layer and IO Layer.Training phase was performed on AWS deep learning AMI.

Model Deployment

  • AWS Sagemaker hosting service

Achievement

  • By implement this solution on cloud based infrastructure. End-user get quick response and reduce the cost for handling infrastructure.
  • User are able to get clear thoughts and visibility for their target product

Technology Stacks

  • Cloud InfrastructureAWS Sagemaker
    AWS Lamda
    Technologies
    Python
    Keras
    Flask Framework
    React
    Postgress SQL
Team
acr

1

React Developer

acr

1

Node Developer

acr

1

Python Developer

acr

1

Cloud Developer

The Result

  • We Successfully developed a robust solution for client’s complex business need to increase the conversion rates and eliminate online shopping returns
  • Created a more personalized customer experience to communicate with their favourite brands in a personal way
  • The Website identify customer behavior and purchase pattern to predict what exactly customers want
  • The team is delighted in assisting the client in carrying out the desired result
Contact us