feat: add Random algorithm to GeneralRecommenders

[guedes-joaofelipe]

A Random algorithm was added as a General Recommender for baseline purposes.

The Random algorithm recommends random items for each user. The motivation behind this approach is to create 2 baselines:

• Coverage metric: a random algorithm tends to have a high Item Space Coverage.
• Ranking/Rating metrics: any recommender algorithm should outperform a random approach in terms o ranking and rating metrics.

The inclusion of the Random class provides easier comparison in terms of the aforementioned metrics

[Paitesanshi]

Thanks for your interest in and contribution to RecBole! I apologize for my delayed response. You are correct that any model should perform better than random. However, I haven't seen any research papers in the field of recommendation systems that use the random method as a baseline for recommendation systems.

[ChenglongMa]

Hi @Paitesanshi, I think random method could be a good baseline when we want to improve performance like diversity, and coverage other than accuracy-based metrics.

Thanks!

[guedes-joaofelipe]

Complementing @ChenglongMa , here is an example of a published paper that explores diversity metrics and uses the Random algorithm as a baseline:

https://dl.acm.org/doi/10.1145/2043932.2043955

[Paitesanshi]

@guedes-joaofelipe @ChenglongMa Thank you for pointing that out. You're right. We've now added the Random recommendation algorithm to RecBole's model library. We appreciate your contribution and feedback to the RecBole！