Epsilon

An application to help online learners find e-learning courses, based on state-of-the-art recommender algorithm.

Comment

Inspiration

  • Lack of recommendation engine on real life Elearning systems.
  • Poor completion rates on typical E-learning systems. Even famous providers as Coursera, Edx,... experiences extremely low rates of completion: ~6.8%
  • Unnecessary time wasted on searching appropriate courses for users.
  • Online courses providers often focus on increasing revenue, not giving good experience for users.

What it does

  • Provide a friendly, portable application with a clear, sensible layout to give personal recommendations for e-learning users. The application is also a channel for providers to reach their target audiences, not only for increasing their sale figures, but also improving learners's experience on their systems.
  • Different types of recommendation method used in the application:
    • Recommend courses for new users based their category preferences.
    • Recommend courses for users based on their history of enrollments.
    • Recommend similar courses for users when making an enrollment.
    • Recommend courses for users based on similar users.

How we built it

  • Research and choose an appropriate algorithm to solve the problem: D. Wu, G. Zhang, and J. Lu, "A Fuzzy Tree Matching-Based Personalized E-Learning Recommender System," IEEE Transactions on Fuzzy Systems., vol. 23, issue. 6, pp. 2412 - 2426, Dec. 2015.
  • Crawl initial data from famous Vietnamese e-learning providers: edumall.vn, kyna.vn, academy.vn
  • Design and implement a complete server system using REST API (NodeJS). The algorithm mentioned above is the heart of the system.
  • Build an Android application to give personal recommendations for e-learning users ## Challenges we ran into
  • The lack of any well-known dataset publicy accessible for research in e-learning recommendation area.
  • Implement a complicated algorithm on a very short time and under a huge pressure of accuracy.
  • Design a friendly UX app for real world e-learning users.

Accomplishments that we're proud of

  • We bring a state-of-the-art algorithm to a real world system, inspite of the lack of appropriate data.
  • We completed a fully-functional system in just 24 hours, a big challenge that we have never experienced before.

What we learned

  • Knowledge about fuzzy logic, data mining, recommendation system architecture, material design...
  • Skills to handle an overwhelming number of problems and make quick decisions.
  • Team-working skill.

What's next for Epsilon

  • Co-operate with MOOCs providers (such as Edumall, Topica Native, Academy.vn, Kyna...) to obtain a reasonable amount of data and help them improve completion rates by targeting users more precisely.
  • Improve accuracy and performance of the core algorithm.
  • Build a web version for the product.

Built With

Share this project:

Updates