Ticket #1178 (closed wish: fixed)
Widget for PCA
|Reported by:||anze||Owned by:||ales|
PCA Widget should have the following
Train data (data table on which the projection is learned)
Transformed data (train data transformed into PC space) Eigen vectors (attribute projection contains eigen vectors in rows)
When widget is opened, it should contain two fields on the left, to specify no. of components and/or variance covered. On the right side it should have a scree plot of the projection showing the variances of components and cumulative variance (similar to scree_plot).
It would be nice if variance / no. components tresholds could be set using the screeplot (similar to how dendrogram works, if only 10 components are selected, a line should separate selected and unselected components in screeplot, this line could be moved to change this treshold
TRANSFORMATION OF ADITIONAL DATASETS
A separate widget should be used for transformation of additional datasets with a learned projection. This widget should basically just apply domain of transformed dataset on the new dataset.