Evaluating Recommendation Algorithms - Offline and Online Approaches

Discover the inner workings of Delivery Hero's recommendation algorithms in this 23-minute video presentation by Manchit Madan. Explore how gradient boosting, NLP, and representation learning are utilized to create personalized suggestions for cart completion and product recommendations. Gain valuable insights into the process of offline and online evaluations for recommender systems, including the importance of A/B testing and strategic offline assessments. Learn how to balance user experience improvements with business metrics, and understand the key factors in selecting the most effective algorithms for various use cases. Delve into topics such as cross-selling recommendations, the necessity of offline evaluation, and the design of online A/B tests, equipping yourself with knowledge to make data-driven decisions in the realm of recommendation systems.

- Introductions
- About the Author
- Cross-selling Recommendations
- Why do we need offline evaluation?
- What is offline evaluation?
- How to Design Online Evaluation A/B Tests?
- Offline decide the best recommendation algorithm
- Conclusion