# Ranking Candidate Books by Three-level Relevance — solution

*English translation by SOTA – AI Community of the Russian original. Organisers who would like this translation removed can email sota.ai.community@gmail.com.*

## Solution

The solution involves training a model that outputs a score for a pair (`user_id`, `book_id`), and then sorting the candidates by decreasing score.

Possible directions for improvement (these are not requirements):

- building features from the interaction history in `train.csv`;
- using the metadata of books and users;
- training a model that distinguishes the strength of the "read" and "planned" signals and cuts off the "cold" candidates.