Back to photostream

Prediction_SqErr_Histogram

This is a plot shows the relative contribution to total (squared) error versus the error's magnitude. The plot shows that errors that are moderate in magnitude contribute most to total squared error (and thus RMSE). Small errors (which are frequent) and large errors (which are infrequent) contribute significantly less, on average.

 

This plot is part of the analysis of algorithms used to predict movie ratings. These algorithms are being used in the Netflix Prize data mining competition (see www.netflixprize.com for more details).

170 views
0 faves
0 comments
Uploaded on December 13, 2007
Taken on December 13, 2007