Example classification models (produced using GP) and the resulting confusion matrices for the Thyroid data set.
Please note: All variables were linearly scaled to the interval [0;100], the threshold values are therefore also values between 0 and 100.
10-fold CV set 0:
class(t) = IF(>=(Log(Log(-(Log([2.643*Var16(t)])|[1.000*Var12(t)])))|
IF(>=(-(-([1.000*Var17(t)]|[1.000*Var16(t)])|+(-0.141724|[1.000*Var18(t)]))|
-(Sig([1.000*Var12(t)])|^([1.000*Var18(t)]| [1.000*Var7(t)])))|ThenElse(-(Sin(
[-2.040*Var16(t)])|Sin([1.973*Var18(t)]))|IF(>=([1.973*Var18(t)] |49.864966)|
ThenElse([2.643*Var16(t)]|[-2.040*Var16(t)])))))|ThenElse(+(-(+(Cos(
[0.427*Var18(t)])|-([1.000*Var17(t)]|1.287233))| Sqrt(+([3.955*Var20(t)]|
[11.739*Var16(t)])))|+(+(*([0.083*Var2(t)]| [0.427*Var18(t)])|Cos(
[0.427*Var18(t)]))|+(*([-0.045*Var20(t)]| [1.000*Var18(t)])|[3.955*Var20(t)])))|
+(+(-2.541577|e^(/([1.973*Var18(t)]| [0.427*Var18(t)])))|Cos(+(Sin(49.864966)|
^([1.000*Var16(t)]|6.173179))))))
Thresholds: [33.45; 75.25]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 137 | 9 | 5 | Est. [0] | 18 | 3 | 0 |
Class [1] | 9 | 337 | 14 | Class [1] | 0 | 15 | 0 |
[2] | 2 | 2 | 5965 | [2] | 0 | 2 | 682 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.37% | 99.31%
10-fold CV set 1:
class(t) = *(+(*(-(-(Cos([1.000*Var16(t)])|Cos([-4.991*Var16(t)]))| *(/(
[1.000*Var7(t)]| 6.767502)|[-0.214*Var16(t)]))|*(*(Sqrt([1.000*Var16(t)])|
3.886609)|+(Cos([1.000*Var16(t)])|2.302652)))|-(+(/(+([1.000*Var16(t)]|
[1.000*Var2(t)])|e^([-0.214*Var16(t)]))|-(+([-4.991*Var16(t)]|
108.231865)|+(0.000000| [1.000*Var2(t)])))|-(*(/(2.302652|1.457826)|+(
[1.000*Var16(t)]|3.886609)) |*(*([-4.991*Var16(t)]|[-0.214*Var16(t)])|
Sqrt([1.000*Var16(t)])))))|IF(<=(-(+(Sqrt([1.000*Var7(t)])|+
([1.000*Var11(t)]|[8.693*Var20(t)]))|+(-(0.000000|[1.006*Var20(t)])|
-(92.692883|0.000000)))|-(-(1.457826|+(0.000000| [1.000*Var4(t)]))|/(-(
[-4.991*Var16(t)]|-4.030073)|^([1.000*Var16(t)]| [8.693*Var20(t)]))))|
ThenElse(e^(-(Sin([-4.991*Var16(t)])| [1.000*Var16(t)]))|e^([-0.214*Var16(t)]))))
Thresholds: [39.225; 72.7]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 149 | 0 | 10 | Est. [0] | 15 | 0 | 1 |
Class [1] | 2 | 332 | 45 | Class [1] | 0 | 36 | 1 |
[2] | 0 | 0 | 5942 | [2] | 0 | 0 | 667 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.12% | 99.72%
10-fold CV set 2:
class(t) = +(+(+(IF(>=(+([2.933*Var5(t)]|[4.635*Var20(t)])|^([1.000*Var16(t)]|
32.512281))|ThenElse(104.077790|+(-5.259905|[3.901*Var20(t)])))|IF(>=(+(
[1.000*Var17(t)]|[1.000*Var17(t)])|+ ([1.000*Var16(t)]|[4.635*Var20(t)]))|
ThenElse(+([1.000*Var2(t)]| [4.635*Var20(t)])|-4.370451)))|IF(>=(^(/(
[1.000*Var16(t)]| [1.000*Var17(t)])|+([1.000*Var8(t)]|[0.373*Var7(t)]))|
*(Sqrt([4.635*Var20(t)])|Log([1.000*Var16(t)])))|ThenElse([1.000*Var16(t)]|
+(+(-5.259905|-4.370451)|+(-5.259905|[0.611*Var7(t)])))))|IF(>=(+(/(+(
[1.000*Var10(t)]|0.000000)|/([1.000*Var16(t)]|[1.000*Var2(t)]))|+(Sqrt(
[2.880*Var20(t)])|+([1.000*Var4(t)]|-4.370451)))|Cos(^(Log([1.000*Var16(t)])|-
(-3.260221|[1.000*Var13(t)]))))|ThenElse(*(^(Sqrt([0.079*Var20(t)])|+(
[1.000*Var14(t)]|-3.260221))|+(+([1.000*Var2(t)]|[1.000*Var17(t)])|+
([1.000*Var17(t)]|[1.000*Var16(t)])))|/(+(+([1.000*Var8(t)]| [0.611*Var7(t)])|+
([1.000*Var8(t)]|[0.373*Var7(t)]))|-(+([2.620*Var20(t)]|[1.000*Var10(t)])|
-(6.267298|[1.000*Var13(t)]))))))
Thresholds: [32.425; 76.475]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 140 | 7 | 2 | Est. [0] | 17 | 0 | 0 |
Class [1] | 5 | 316 | 27 | Class [1] | 4 | 45 | 3 |
[2] | 0 | 0 | 5983 | [2] | 0 | 0 | 651 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.37% | 99.03%
10-fold CV set 3:
class(t) = +(+(IF(<=([-0.37921775*Var16(t)]|+(-0.43533937|
[-2.76239437*Var7(t)]))|ThenElse( +([2.64778823*Var7(t)]|
-0.91019582)|[1.26036869*Var19(t)]))|IF(OR(>=(-0.29808521|-0.96359298)|
>=([1.88284005*Var12(t)]|[0.14167759*Var6(t)]))| ThenElse([0.16761998*Var2(t)]|
-0.73643814))) |+([0.82477057*Var17(t)]|+
([-0.31729187*Var16(t)]|+(+([3.68276008*Var20(t)]|[-0.04266986*Var17(t)])|
-(4.50479669|2.31342641)))))
Thresholds: [43.125; 89.025]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 136 | 4 | 10 | Est. [0] | 16 | 1 | 0 |
Class [1] | 9 | 331 | 231 | Class [1] | 3 | 28 | 23 |
[2] | 2 | 3 | 5754 | [2] | 0 | 1 | 648 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 96.00% | 96.11%
10-fold CV set 4:
class(t) = IF(<=(/(-(*(-0.507917|[1.000*Var8(t)])|[0.875*Var16(t)])|-(
[0.253*Var2(t)]|Log([0.875*Var16(t)])))|+(*(*(+([1.000*Var4(t)]|
[0.720*Var7(t)])|Sin([0.875* Var16(t)]))|/([-2.651*Var16(t)]|-(
[0.875*Var16(t)]|[1.000*Var20(t)])))|+([-0.548*Var16(t)]|e^(e^(
[9.284*Var12(t)])))))|ThenElse(-(-(-(-(0.041530|-100.301158)|0.000000)|
Sin(12.938045))|/(/([-2.651*Var16(t)]|3.143818)|-([9.284*Var12(t)]|
[1.000*Var20(t)])))|+(+(*(+([1.000*Var2(t)]|[0.720*Var7(t)])|Sin(12.938045))|+
(*([0.012*Var17(t)]|[1.000*Var17(t)])|[-0.548*Var16(t)]))|-(-(-([5.281*Var20(t)]
|[0.856*Var20(t)])|[0.856*Var20(t)])|+(Cos([0.780*Var20(t)])|-([0.856*Var20(t)]
|[1.000*Var17(t)]))))))
Thresholds: [35.55; 77.95]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 134 | 8 | 5 | Est. [0] | 15 | 4 | 1 |
Class [1] | 13 | 309 | 33 | Class [1] | 1 | 46 | 4 |
[2] | 1 | 1 | 5976 | [2] | 2 | 0 | 647 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.06% | 98.33%
10-fold CV set 5:
class(t) = ^(IF(>=(+(+(^(90.826907|[1.350*Var16(t)])|+([1.350*Var16(t)]|
[1.000*Var3(t)]))|-(Cos([1.000*Var9(t)])|*([1.000*Var5(t)]|
[0.428*Var20(t)])))|^(^(+([1.000*Var2(t)]|98.567908)|Cos([-0.239*Var16(t)]))
|+(+([1.000*Var12(t)]|[1.000*Var4(t)])|+([1.000*Var6(t)]|1.584749))))|
ThenElse(+(-(+(-3.514523|[3.551*Var20(t)])|-([0.428*Var20(t)]|
[1.000*Var17(t)]))|IF(<=([1.813*Var7(t)]|[2.014*Var16(t)])|
ThenElse([0.442*Var2(t)]|60.781038)))|+(98.567908|-(-(1.584749|[-0.239*Var16(t)])|
^([1.000*Var16(t)]|1.584749)))))|Sig(+(^(+(+(
[2.496*Var20(t)]|0.000000)|+([-0.239*Var16(t)]|0.000000))|Cos(55.840663))|
*(Cos(-([1.000*Var9(t)]|[0.428*Var20(t)]))|
*(1.193392|+([1.350*Var16(t)]|[1.350*Var16(t)]))))))
Thresholds: [35.65; 79.5]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 145 | 4 | 5 | Est. [0] | 15 | 1 | 1 |
Class [1] | 5 | 326 | 29 | Class [1] | 1 | 34 | 5 |
[2] | 0 | 3 | 5963 | [2] | 0 | 0 | 663 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.29% | 98.89%
10-fold CV set 6:
class(t) = IF(<=(IF(<=(+([0.306*Var7(t)]|e^([1.000*Var12(t)]))|-(-(
[4.358*Var20(t)]|0.587528)|Sqrt([0.500*Var16(t)])))|ThenElse(Cos(/
([1.000*Var4(t)]|[1.273*Var17(t)]))|[1.000*Var4(t)]))|IF(<=(+(
[7.775*Var20(t)]|+([1.273*Var17(t)]|[ 2.186*Var20(t)]))|+(-(0.000000|
-5.218301)|98.760572))|ThenElse([1.000*Var8(t)]|IF(>=(0.000000|
[1.000*Var2(t)])|ThenElse([0.306*Var7(t)]|[7.775*Var20(t)])))))|
ThenElse(IF(<=(+([0.500*Var16(t)]| [1.000*Var12(t)])|-(-([4.358*Var20(t)]|
0.587528)|[1.000*Var12(t)]))|ThenElse(+(+(0.587528|98.760572)|0.587528)|
[1.000*Var12(t)]))|IF(<=(+([0.500*Var16(t)]|e^([2.353*Var16(t)]))|+(
Sqrt([1.000*Var4(t)])|15.374653))|ThenElse(+(+(0.587528|98.760572)|0.587528)|+(+(
[1.273*Var17(t)]| [2.186*Var20(t)])|/ ([3.649*Var18(t)]|9.638362))))))
Thresholds: [36.425; 80.1]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 145 | 9 | 4 | Est. [0] | 11 | 0 | 1 |
Class [1] | 8 | 325 | 26 | Class [1] | 1 | 33 | 1 |
[2] | 0 | 0 | 5963 | [2] | 1 | 1 | 671 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.27% | 99.31%
10-fold CV set 7:
class(t) = +(^(+(+(+(Log(8.775252)|+([1.440*Var20(t)]|8.775252))|*(Sin(
[0.158* Var20(t)])|-(8.775252|[1.000*Var16(t)])))|e^(Sin(+([0.158*Var20(t)]|
[1.000*Var16(t)]))))|Sig(+(+(Sin([-0.421*Var16(t)])|+([0.086*Var4(t)]|
[1.285*Var20(t)]))|+(+([-1.893*Var16(t)]|[1.000*Var8(t)])|
Sin([1.000*Var16(t)])))))|IF(<=(Sig(+(-(59.824802|[1.000*Var2(t)])|
-(8.775252|[1.000*Var19(t)])))|+(Log(+([1.000*Var2(t)]|8.775252))|+(e^(
[1.000*Var7(t)])|+([-1.893*Var16(t)]|0.000000))))|ThenElse(+(59.824802|
-(2.810579|[0.158*Var20(t)]))|+(*(*([0.249*Var2(t)]|[0.158*Var20(t)])|
[0.158*Var20(t)])|+(*([0.158*Var20(t)]|[0.158*Var20(t)])|
+([0.086*Var4(t)]|[1.000*Var17(t)]))))))
Thresholds: [41.575; 74.575]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 140 | 11 | 5 | Est. [0] | 9 | 3 | 3 |
Class [1] | 13 | 305 | 30 | Class [1] | 3 | 49 | 4 |
[2] | 1 | 0 | 5975 | [2] | 0 | 0 | 649 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.07% | 98.19%
10-fold CV set 8:
class(t) =
-(IF(<=(-(+(+([5.748*Var20(t)]|[1.000*Var4(t)])|3.807626)|*(
-(1.928774|[1.000*Var7(t)])|-([3.224*Var16(t)]|[1.000*Var8(t)])))|
IF(<=(-([3.351*Var16(t)]|[1.000*Var2(t)])|*(1.000000|3.807626))|
ThenElse([-5.157*Var20(t)]|+([4.698*Var16(t)]|56.916803))))|
ThenElse(-(+(+([1.000*Var16(t)]|-8.636405)|+([5.518*Var16(t)]|-8.636405))|
-(-([-5.157*Var20(t)]|[1.000*Var16(t)])|e^([0.222*Var17(t)])))|
IF(<=(-([3.351*Var16(t)]|[1.000*Var7(t)])|*(1.000000|3.807626))|
ThenElse(+([5.518*Var16(t)]|102.747052)|+([1.000*Var20(t)]|56.916803)))))|
IF(>=(-(3.807626|+(+([1.000*Var15(t)]|[1.000*Var11(t)])|+([1.000*Var7(t)]|
[-5.157*Var20(t)])))|Log(/(e^(56.916803)|Sqrt([4.697*Var2(t)]))))|
ThenElse(-(-(3.807626|*([3.224*Var16(t)]|5.880540))|Cos(+([4.698*Var16(t)]
|1.000000)))|*(/(+(3.464115|[1.000*Var4(t)])|+([1.000*Var4(t)]|1.928774))|+(
[4.698*Var16(t)]|Sig([5.748*Var20(t)]))))))
Thresholds: [36.375; 83.275]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 143 | 11 | 3 | Est. [0] | 16 | 1 | 3 |
Class [1] | 6 | 317 | 33 | Class [1] | 0 | 38 | 6 |
[2] | 1 | 1 | 5965 | [2] | 0 | 0 | 656 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.15% | 98.61%
10-fold CV set 9:
class(t) = +(+(IF(>=(^(Log([3.598*Var20(t)])|+([2.767*Var7(t)]|[1.000*Var12(t)]))|
Log([2.368*Var16(t)]))|ThenElse(IF(>=([2.368*Var16(t)]|35.377930)|
ThenElse(-21.656405|28.887860))|+(+([1.000*Var17(t)]|-3.329153)|+
([0.247*Var2(t)]|-3.329153))))|IF(>=(+(+([1.000*Var6(t)]|28.887860)|+
([2.368*Var16(t)]|[1.000*Var10(t)]))|*(-([0.861*Var16(t)]|[1.000*Var3(t)])
|[1.000*Var18(t)]))|ThenElse(IF(>=([2.368*Var16(t)]|49.112689)|ThenElse(
22.857204|49.112689))|-([3.484*Var18(t)]|+([1.000*Var18(t)]|-5.595520)))))|
+(IF(<=(-([2.368*Var16(t)]|/(1.182745|[1.000*Var17(t)]))|
+([1.000*Var8(t)]|[1.000*Var17(t)]))|ThenElse(Sin(Sig(-21.656405))|
IF(<=(35.377930|[3.722*Var20(t)])|ThenElse([0.894*Var16(t)]|-21.656405))))|
IF(>=(^(-([1.000*Var18(t)]|[1.000*Var2(t)])|+([2.279*Var7(t)]|
[1.000*Var4(t)]))|Log([2.368*Var16(t)]))|ThenElse(IF(>=([0.861*Var16(t)]|
22.857204)|ThenElse([2.279*Var7(t)]|22.857204))|
IF(<=(35.377930|[3.598*Var20(t)])|ThenElse([0.247*Var2(t)]|-21.656405))))))
Thresholds: [34.225; 75.6]
TRAINING TEST
Orig. -> | [0] | [1] | [2] | Orig. -> | [0] | [1] | [2] |
----------|-----|-----|------| ----------|-----|-----|-----|
Est. [0] | 145 | 0 | 3 | Est. [0] | 17 | 0 | 1 |
Class [1] | 3 | 341 | 21 | Class [1] | 1 | 27 | 6 |
[2] | 0 | 0 | 5967 | [2] | 0 | 0 | 668 |
----------|-----|-----|------|------- ----------|-----|-----|-----|-------
| 99.58% | 98.89%