time-SVD++ is now implemented in Graphlab

One of the algorithms to perform better in the KDD CUP contest this year is Koren's time-SVD++ algorithm. You can find more details in a previous post.

To summarize, according to our testing, we got the following validation RMSE on KDD CUP data:
(as lower prediction error, the better..)

1 Neighborhood model Adjusted cosine (AC) similarity 23.34
2 ALS                                                                              22.01
3 Weighted ALS                                                              18.87**
4 BPTF                                                                             21.84
5 SGD                                                                              21.88
6 SVD++                                                                         21.59
7 Time aware neighborhood                                             22.7
8 time-SVD                                                                      21.41
9 time-SVD++                                                                 20.90
10 MFITR                                                                        21.30
11 time-MFITR                                                                21.10
12 Random forest                                                              26.0


(Weighted ALS did not perform well on test data).
As can be seen from the above table, time-SVD++ was the best performing single algorithm
on the KDD CUP data. I also got the same impression when talking to Yehuda Koren.

time-SVD++ is now implemented as part of GraphLab's collaborative filtering library.
You are all welcome to try it out!