SUPPORT VECTOR MACHINE

chardfriendlyΤεχνίτη Νοημοσύνη και Ρομποτική

16 Οκτ 2013 (πριν από 4 χρόνια και 23 μέρες)

79 εμφανίσεις

SUPPORT VECTOR MACHINE



============ Project Parameters ============


Target variable: SNB

Number of predictor variables: 8

Type of model: Support Vector Machine (SVM)

Type of SVM model: C
-
SVC

SVM kernel function: Radial Basis Function (RBF)

Type of an
alysis: Classification

Category weights (priors): Specified values

Misclassification costs: Probability threshold = 0.200000

Validation method: Cross validation

Number of cross
-
validation folds: 10



============ Input Data ============


Number of varia
bles (data columns): 9

Data subsetting: Use all data rows

Number of data rows: 1132

Total weight for all rows: 1132

Rows with missing target or weight values: 0

Rows with missing predictor values: 0



============ Summary of Variables ============



Var
iable Class Type Missing rows Categories

-----------

---------

-----------

------------

----------

Sex Predictor Categorical 0 2

Age Predictor Continuous 0 71

Clark Pred
ictor Categorical 0 4

Breslow Predictor Continuous 0 50

Regression Predictor Categorical 0 2

Ulceration Predictor Categorical 0 2

Hystotype Predictor Categorica
l 0 3

Mitotic index Predictor Categorical 0 3

SNB Target Categorical 0 2





============ SVM Parameters ============


Type of SVM model: C
-
SVC

SVM kernel function: Radial Basis

Function (RBF)


SVM grid and pattern searches found optimal values for parameters:


Search criterion: Minimize total error


Number of points evaluated during search = 150


Minimum error found by search = 0.223204


Parameter values:


C = 2.003204


Gam
ma = 0.003190


Number of support vectors used by the model = 1011




============ Misclassification Tables ============



---

Training Data
---



--------
Actual
--------

-------------
Misclassified
-------------


Category Count Weigh
t Count Weight Percent Cost


--------

--------

------------

--------

------------

-------

------


0 900 900 487 487 54.111 0.541


1 232 232 29 29

12.500 0.125


--------

--------

------------

--------

------------

-------

------


Total 1132 1132 516 516 45.583 0.456



---

Validation Data
---



--------
Actual
--------

-------------
Misclassi
fied
-------------


Category Count Weight Count Weight Percent Cost


--------

--------

------------

--------

------------

-------

------


0 900 900 583 583 64.778 0.648


1

232 232 24 24 10.345 0.103


--------

--------

------------

--------

------------

-------

------


Total 1132 1132 607 607 53.622 0.536




============ Confusion Matrix ===
=========




--------

Training Data
--------



Actual :
----
Predicted Category
---

Category: 0 1

--------
:
------------

------------


0: 413 487


1: 29 203




--------

Validation Data

--------



Actual :
----
Predicted Category
---

Category: 0 1

--------
:
------------

------------


0: 317 583


1: 24 208




============ Overall Importance of Variables ============



Variable Importance

-----------

----------

Mitotic index 100.000

Regression 34.483

Clark 25.862

Breslow 15.517

Histotype 8.621

Sex 3.448

Ulcerat
ion 1.724