Recommendation and Classification System for Online Learning Platform

iconEdTech
iconData Science
iconMachine Learning

We enhanced the leading European online learning platform with deep learning solution to ensure flawless customer experience. Majority of content watched online comes from recommendations: 80% on Netflix, and 60% on YouTube, as per the research (link to external resource). So one of the popular massive European online learning aggregating platforms opted for a robust recommendation and classification system. Read on to find out how we beat such challenges.

icon
Challenge

Our client aimed to increase engagement metrics and user’s lifetime value for their platform. We took on the challenge to improve the client's customer experience. Also, new functionality should automatically classify any new online course added, tailored to any user preferences. See what’s under the hood of solutions like this one.

icon
Solution
  • First, our data science professional built a data collection mechanism. For this purpose, we’ve designed a data storage and cleaned data for further steps.
  • Afterward, we designed a recommender system tailored per learner’s preferences relying on natural language processing (NLP) techniques. We opted for content-based filtering.
  • Developing classification system for content added on the rolling bases.
  • Integrating recommender and classification systems into learning platform.
icon
Impact

We improved customer experience by increasing engagement metrics by 18% and user's LTV by 13%, which has a significant impact on income equivalent for our client. Moreover, as the platform operates as an aggregator, all the content from any new component added could be geared to the user's personal preferences.

icon
Tech Stack

Python

REST

Ubuntu

Docker

SQL Server

Frameworks and Libraries

numpy

pandas

matplotlib

seaborn

scikit-learn

scipy

implicit

cornac

recommenders

Flask

gunicorn

AI Development Stack

Data exploration

Data cleansing

Modeling

Deployment (on-premise server)

So this is how the combination of NLP and classification modeling contributed to our client's flawless learning experience and higher income.

astronaut