Recommendation is all around us, whether you are watching YouTube, scrolling Facebook, Instagram, Tiktok, or buying stuff on Amazon. It improves our life by showing us relevant, personalised items. But building a recommendation system is not straightforward. The recommendation system is a fast-changing field; knowing which recommender algorithms and system architectures work for your business is a vital skill.
I have been working in the recommendation field for over 5 years and was fortunate to participate in building core recommendation systems. I will condense all of my knowledge into a series of articles. This first article will give you an overview of the recommendation system.
Note: My expertise is in e-commerce recommendation. It may have some caveats when we apply it to other areas.
The Objective
Having a clear objective of what your recommendation system needs to do is essential for your business. In general, we can categorise our objective into 4 parts:
- Item-to-Item (I2I): Given a target item, what other relevant things can we show to users? For example, when we click on a YouTube video, we can see other relevant videos recommended. Or when we view an item on Amazon, other relevant products are also suggested.
