Showing Predictor "NACD (L1) -L2ML"

This predictor was successfully trained!

Meta

Notes:

Like http://pssp.srv.ualberta.ca/predictors/32, but using L2-marginal loss for optimization.

(Marginal Loss Type: weighted)

Publicly Accessible: yes

Viewable by Administrators: yes

Parent Predictor: Show

Actions:Edit Delete

Hide

Dataset Show

Options
Regularization: L1	Tuning Score: avg marginal l1-loss
Time Points: 50	Tune Parameters: true
Feature Selection: false	Include Sibling Features: false
Standardize: true	Run Cross-Validation: true
Objective Function: l2 marginal loss	Marginal Loss Type: weighted
Use pre-defined Fold: No

Results
Start Time: 2016-07-28 08:14:16 -0600	Used Time: 18 minutes
Label Name: SURVIVAL	C Parameter: 1000.0
PSSP model: mtlr_model_file2180.mtlr	Version: 0.0.1
Feature Sparsity Index: 46.1% used

Predictor Settings / Retrain Show

Cross-validation results

5-Fold Cross-Validation Statistics*
Measure	PSSP Predictor (median)	PSSP Predictor (mean)	K-M Predictor
Concordance Index	0.73 ± 0.0		0.5 ± 0.0
Hinged L1 Loss	11.34 ± 0.58		13.28 ± 0.05
Uncensored L1 loss	13.69 ± 1.6		12.18 ± 0.05
Marginal L1 Loss	18.02 ± 0.38		20.96 ± 0.16
Marginal L2 Loss	22.96 ± 0.36		27.14 ± 0.29
Hinged L1 Log-Loss	0.65 ± 0.01		0.79 ± 0.01
Uncensored L1 Log-Loss	0.9 ± 0.04		0.97 ± 0.01
Marginal L1 Log-Loss	0.83 ± 0.01		1.02 ± 0.01
Log-Likelihood Loss	2.9 ± 0.05
D-calibration χ² statistics	96.88 ± 19.09		0.97 ± 0.12
D-calibration χ² p-value	0.0 ± 0.0		1.0 ± 0.0

*Mean value ± standard deviation.

Examine Classification Accuracy