I would like to know which method SVD or KNN will yield better prediction accuracy in recommendation systems. Has anyone done a comparative study which I can refer to.
It depends on the problema. I suggest to employ both. Under these conditions, it is posible to see how they perform over your particular problem
Generally, SVD provides more accurate prediction compare with KNN.
KNN method is based on k nearest neighbor users or items followed by top K users and items is chose.
KNN can be applied on raw data or on lower dimensions of the processed data.
As Erik mentioned it depends on the problem.
pure SVD is not useful for prediction.
pure SVD when applied on prepossessed data gives better results with lower dimensions.
Low rank approximation methods combined with matrix factorization techniques gives better results.
check the following papers
- An Effective Preprocessing Algorithm for Model Building in Collaborative Filtering based Recommender System.
- A Scalable Ensemble Architecture for Collaborative Filtering in Recommender Systems
The accuracy depends on the source of the data and the target objetive.
Most of cases, kNN is better in sets of data with low missing date proportion. And SVD is better in huge size data sets.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs