Hi everyone,
In the past few months, I've been working to build ML-based recommender systems for Steam video games. I built this since I'm not satisfied with recommendations shown by Steam, especially for non-popular games.
SRec provides 3 kinds of recommendation systems,
* Smart recommender. It's based on GCE-GNN (Global Context Enhanced Graph Neural Networks) model with some modifications.
* Recommendation by similar game tags. This recommender provide explanation by showing top-5 most similar tags between chosen and recommended games.
* Recommendation by Steam user preferences.
It's strongly recommended to use either smart recommender or recommendation by game tags. Currently, smart recommender has poor performance when recommending extremely popular games. The performance of recommendation by similar game tags is limited by how Steam user apply tags to each games. Some technical details of Smart recommender can be seen on https://srec.ai/blog/srec-gcegnn.
On each game page, SRec shows various game details including Steam Deck compatibility, content warning and all game tags. SRec also shows review insight for most games that shows most frequently mentioned keywords where you can read some top reviews which mention selected keyword.
While you can access SRec on either desktop or mobile device, I recommend you to access SRec with desktop since I have no experience building front-end. But feel free to ask any question or leave feedback.
