Is Supervised Learning Better than Unsupervised Learning (For Recommendation Systems)

Question

I am working on a Recommendation System as a personal project (I finish it on time, I'll present it as my final year project). I devised mathematical methods that estimated User's ratings of products based on their previous ratings, without using ML. I developed a host of such algorithms for recommendation (I developed a paradigm that allowed me to generate an infinite number of distinct recommendation systems). I intended to assign weights to a select few of them (maybe the ones with the highest accuracy), and take the weighted average of the estimated user ratings.

I realised that Machine Learning could be used to best determine the appropriate weight to assign each of the algorithms on a per user basis, and decided to start learning ML yesterday. It seems what I'm intending to do, is supervised learning.

From what I can tell, it seems existing recommendation systems on the market use unsupervised learning. I am curious to how the two paradigms compare in my case. Would supervised learning work better than unsupervised learning, or is unsupervised learning a better paradigm for recommendation systems?

Answer 1

It doesn't make sense to ask if supervised learning is better or worse than unsupervised since they are used in different contexts.

In the most simple terms, supervised learning is used when you have a dependent variable, so you're investigating the relationship between this dependent variable and one or more independent variables. Often you're trying to predict something about the dependent variable using the independent variable(s).

In unsupervised learning, you are essentially looking for certain kinds of structure in the data or trying to find various kinds of relationships between variables, but none are considered special in the sense of being a dependent variable.