Hybrid Recommender

A weighted blend of content-based and collaborative approaches that covers each one's weaknesses: r^=α⋅r^CB+β⋅r^Item-CF\hat{r} = \alpha \cdot \hat{r}_{\text{CB}} + \beta \cdot \hat{r}_{\text{Item-CF}} with α+β=1\alpha + \beta = 1.

When to use

Worked example from the lecture: a content-based prediction of 2.8 and an item-CF prediction of 2.5, with weights 0.4 and 0.6, give 2.62. Raise the content-based weight for new users and the CF weight for active users. The hybrid had the best results in the Seret+ case.

See Lec 6.

Appears in

No references yet.