4
§
PROC DMNEURL: Approximation to PROC NEURAL
------------ Optimization Cycle (Stage=0) --------------
------------ Activation= SQUARE (Stage=0) --------------
NOTE: ABSGCONV convergence criterion satisfied.
SQUARE: Iter=5 Crit=0.06782364: SSE=808.457819 Acc= 81.6443
------------ Activation= TANH (Stage=0) ----------------
NOTE: ABSGCONV convergence criterion satisfied.
TANH: Iter=4 Crit=0.06802595: SSE=810.869323 Acc= 81.6275
------------ Activation= ARCTAN (Stage=0) --------------
NOTE: ABSGCONV convergence criterion satisfied.
ARCTAN: Iter=5 Crit=0.06795346: SSE=810.005204 Acc= 81.6611
------------ Activation= LOGIST (Stage=0) --------------
NOTE: ABSGCONV convergence criterion satisfied.
LOGIST: Iter=11 Crit=0.06802943: SSE= 810.91085 Acc= 81.6107
------------ Activation= GAUSS (Stage=0) ---------------
NOTE: ABSGCONV convergence criterion satisfied.
GAUSS: Iter=10 Crit=0.07727582: SSE=921.127726 Acc= 80.2517
------------ Activation= SIN (Stage=0) -----------------
NOTE: ABSGCONV convergence criterion satisfied.
SIN: Iter=5 Crit=0.06811774: SSE= 811.96345 Acc= 81.6611
------------ Activation= COS (Stage=0) -----------------
NOTE: ABSGCONV convergence criterion satisfied.
COS: Iter=9 Crit=0.07419096: SSE=884.356261 Acc= 81.1913
------------ Activation= EXP (Stage=0) -----------------
NOTE: ABSGCONV convergence criterion satisfied.
EXP: Iter=9 Crit=0.06798656: SSE= 810.39974 Acc= 81.5436
The following approximate accuracy rates are based on the discrete values of the
predictor (
Ü
) variables:
Approximate Goodness-of-Fit Criteria (Stage 0)
Run
Activation
Criterion
SSE
Accuracy
1
SQUARE
0.067824
808.457819
81.644295
3
ARCTAN
0.067953
810.005204
81.661074
8
EXP
0.067987
810.399740
81.543624
2
TANH
0.068026
810.869323
81.627517
4
LOGIST
0.068029
810.910850
81.610738
6
SIN
0.068118
811.963450
81.661074
7
COS
0.074191
884.356261
81.191275
5
GAUSS
0.077276
921.127726
80.251678
After running through the data set we obtain the correct accuracy tables:
Classification Table for CUTOFF = 0.5000
Predicted
Activation
Accuracy
Observed
1
0
Purpose of PROC DMNEURL
§
5
SQUARE
81.610738
1
229.0
960.0
0.067548
0
136.0
4635.0
TANH
82.063758
1
254.0
935.0
0.067682
0
134.0
4637.0
ARCTAN
81.761745
1
242.0
947.0
0.067722
0
140.0
4631.0
LOGIST
81.845638
1
221.0
968.0
0.067818
0
114.0
4657.0
SIN
81.275168
1
222.0
967.0
0.067867
0
149.0
4622.0
EXP
81.543624
1
197.0
992.0
0.068101
0
108.0
4663.0
COS
81.359060
1
101.0
1088.0
0.073967
0
23.0000
4748.0
GAUSS
80.167785
1
7.0000
1182.0
0.079573
0
0
4771.0
The activation function SQUARE seems to be most appropriate for the first stage
(stage=0) of estimation. However, TANH yields an even higher accuracy rate:
Goodness-of-Fit Criteria (Ordered by SSE, Stage 0)
Run
Activation
SSE
RMSE
Accuracy
1
SQUARE
805.19026
0.367558
81.610738
3
ARCTAN
805.89106
0.367718
81.778523
8
EXP
806.66533
0.367895
81.593960
4
LOGIST
807.30313
0.368040
81.778523
2
TANH
807.72088
0.368135
81.778523
6
SIN
809.31533
0.368499
81.291946
7
COS
881.68579
0.384622
81.359060
5
GAUSS
949.21059
0.399078
80.167785
The following is the start of the second stage of estimation (stage=1). It starts with
selecting three eigenvectors which may predict the residuals best:
Component Selection: SS(y) and R2 (Stage=1)
Comp
Eigval
R-Square
F Value
p-
Value
23
4763.193233
0.023292
142.109442
<.0001
21
5192.070258
0.018366
114.178467
<.0001
24
4514.317020
0.017493
110.756118
<.0001
When fitting the first order residuals the average value of the objective function
dropped from 0.068 to 0.063. For time reasons the approximate accuracy rates are
not computed after the first stage: