MoogleLabs Case Study | Skincare Recommendation System

Overview

We worked on creating an ingredient recommender system for an e-commerce website that wanted to improve its user experience.

They wanted to utilize Machine Learning to provide the best recommendations to every client, as per different attributes, to improve results and get repeat customers.

Our Contribution

A leading beauty e-commerce store that offers multiple brands wanted to provide personalized recommendations as per users' skin concerns. So, we created a custom recommender system using AI and ML that suggests the ingredients to use per the client's features, like skin type, texture, etc.

What did we do?

To build the recommendation system, we first selected credible sources of reviews, where we could also find information related to the reviewers' profiles, including skin type, texture, and other important information.

Then, we exported the reviews in a CVS file and ran a sentiment analysis using Natural Language Processing and Natural Language ToolKit. It showed us whether the overall sentiments were positive or negative.

Also, the upvotes and downvotes on the product helped us determine how strong the sentiments were among the general public.

We also extracted the list of ingredients for each product, combined with other data, to determine the components worked best for specific skin types and textures with the help of machine learning.

To find relevant information related to each ingredient found in the products, we relied on scientific websites with high credibility to fetch data.

Tools Used to Create The
Machine Learning

Natural Language Processing

Data

Natural Language Toolkit

Our Process

After collecting all the data in CVS files, we created numerous segregations to extract insights from the data and created the perfect database for machine learning. Now, depending upon the user's input, the machine can offer recommendations of ingredients identified and categorized to yield results for their skin and texture.

The resultant product was deployed on the client's website as a tool that customers could use to avail custom recommendations per their needs.

About the Business

Our client offers skincare products through an eCommerce site on the internet and physical stores.

They have a range of products with various ingredients that are marketed for their potent ingredients.

Requirements

With the lockdown underway, more users were shopping online, and they wanted a system that could offer bespoke recommendations like their experts in the physical stores while limiting cost.

What Problem Arose & Our Solutions?

Problem 1

Finding the correct sites

Solution

Any machine learning system is as good as its raw data. So, we only took data from the most credible sights.

Problem 2

Lack of Insights

Solution

While the client’s websites and a few others did offer essential data, there was a lack of insights. So, we had to create several segregations, quantify people’s perspectives using NLP and draw appropriate insights.

Problem 3

How to Extract User’s Requirements

Solution

To understand users’ requirements, we have created a survey that asks relevant questions and only offers recommendations.

The Results

A highly scalable skincare recommendation system that takes input from users to provide ingredient suggestions. It can also be applied to several other industries with only a few tweaks in data.

Creating the tool has helped the organization make bespoke recommendations, ensuring better results, and getting more repeat customers.

Artificial Intelligence

DevOps

Machine Learning

Consulting

Blockchain

Metaverse

IoT

Data Analytics

AI & ML Services –

Overview

Our Contribution

What did we do?

Tools Used to Create The
Machine Learning

Our Process

About the Business

Requirements