I plan to design a recommendation system, especially with Scikit-Surprise.
A bit of background:

  1. I want to recommend products to shop.
  2. Here, the user is the shop and the items are products (water, chocolate, cheese,...).

So to create the scores and the rating matrix, I use a quota calculation that I normalize to get a score between 0 and 1, where 1 means that the store is the biggest seller of this product(in percentage) (all proportion kept).

First question : is it a good scoring ? is it a good idea to have a rating between 0 and 1 ?

Second question, I want my system to take into account the time of year. For example it recommends more chocolate at Christmas. How can I do that ?
I have an idea where I do "post-filtering", (at the end, when I already compute SVD or system like this).
So in this idea, I try to modify the rating of a product according to the period of the year and according to the evolution of the sales of this product on the sales history. So if the product is very well sold, for example 35% more than the average at this time of year, I increase the score by 35%.
Do you think this is a good idea? Do you have any suggestions for doing this? Do you know some library that can do that easily ?

Thaks for your reading and... stay at home!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.