ML using Decision Trees to Predict NASH from Hepatic NGS GE Profiles
Jim Nelson
April 20, 2015
Goal is to create a model using decision trees to predict the presence of definitive NASH from NGS data of the top 25 most significantly differentialy expressed genes
load packages
install.packages(“randomForest”) install.packages(“gbm”) install.packages(“rpart”) install.packages(“rattle”) install.packages(“rpart.plot”) install.packages(“RColorBrewer”)
read in file
data<- read.csv("./defNASH top 25.csv")
dim(data)## [1] 95 27str(data)## 'data.frame': 95 obs. of 27 variables:
## $ subject : Factor w/ 95 levels "NASH1","NASH10",..: 53 64 75 86 91 92 93 94 95 54 ...
## $ defNASH : int 0 0 0 0 0 0 0 0 0 0 ...
## $ ENSG00000213606: int 0 0 0 3 0 0 10 3 1 4 ...
## $ ENSG00000083857: int 756 492 732 1450 679 1602 1805 541 635 362 ...
## $ ENSG00000198074: int 2 2 1 16 0 8 51 23 10 30 ...
## $ ENSG00000173391: int 4 2 0 1 0 5 4 3 0 0 ...
## $ ENSG00000165912: int 59 84 51 136 222 288 126 124 51 62 ...
## $ ENSG00000115009: int 0 2 0 3 0 0 3 1 1 0 ...
## $ ENSG00000148773: int 7 9 5 42 25 58 43 25 17 7 ...
## $ ENSG00000129474: int 84 128 75 373 326 450 347 241 187 127 ...
## $ ENSG00000138448: int 191 212 338 687 548 977 1050 521 334 270 ...
## $ ENSG00000169429: int 6 0 0 0 2 1 5 1 1 1 ...
## $ ENSG00000156535: int 10 23 24 52 37 85 56 40 24 24 ...
## $ ENSG00000196177: int 2052 2777 2825 4190 9494 10358 6260 5327 1651 2406 ...
## $ ENSG00000060982: int 10 18 10 33 16 44 65 15 18 9 ...
## $ ENSG00000234964: int 2 2 0 6 0 3 0 2 3 2 ...
## $ ENSG00000187498: int 138 200 94 438 391 755 648 264 251 106 ...
## $ ENSG00000231991: int 69 57 85 162 78 201 353 105 121 63 ...
## $ ENSG00000078098: int 1 1 2 5 4 7 6 5 3 2 ...
## $ ENSG00000024526: int 1 1 0 2 1 11 3 0 1 2 ...
## $ ENSG00000079931: int 3 5 2 11 2 12 6 6 11 0 ...
## $ ENSG00000137161: int 57 102 131 187 239 400 230 208 79 130 ...
## $ ENSG00000173744: int 117 158 189 368 357 635 573 250 159 177 ...
## $ ENSG00000118785: int 8 29 4 106 34 92 81 65 48 10 ...
## $ ENSG00000182718: int 122 55 87 152 103 176 271 111 73 64 ...
## $ ENSG00000162366: int 1 1 2 3 3 1 1 2 1 1 ...
## $ ENSG00000168542: int 567 786 631 2166 1599 2575 3674 2085 944 704 ...factorize the dependent variable
data$defNASH <- as.factor(data$defNASH)
data$subject <- NULLdefine the train and test set
NumbofObs <- dim(data)
SplitofTrainTest <- 0.5
train <- sample(1:NumbofObs, SplitofTrainTest*NumbofObs)## Warning in 1:NumbofObs: numerical expression has 2 elements: only the first
## usedtest <- -train
TrainSet <- data[train, ]
TestSet <- data[test, ]
Testing.Outcome <- data$defNASH[test]Set seed for reproducibility
set.seed(123)Recursive Partitioning Models
Build the decision tree
rpart.model <- rpart(defNASH ~ ., TrainSet, method = "anova")Visualize the decision tree using rpart.plot
fancyRpartPlot(rpart.model) 
Make prediction using the test set
predict.rpart <- predict(rpart.model, newdata = TestSet)Vizualize the prediction results
rpart.results <- table(observed = Testing.Outcome, predict = predict.rpart)
rpart.results #confusion matrix## predict
## observed 1.15789473684211 1.71428571428571 1.75 2
## 0 12 4 5 2
## 1 8 3 3 11summary(rpart.results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 7.603, df = 3, p-value = 0.05496
## Chi-squared approximation may be incorrectAs shown by the Chi2 and pval the model worked ok but lets see how random Forest will perform
Bagging Method
Bagging is same as Random Forest but with full feature space (m = p)
Build the decision tree
bag.model <- randomForest(defNASH ~ ., TrainSet, mtry= 25, importance= T )Vizualize the model results
bag.model##
## Call:
## randomForest(formula = defNASH ~ ., data = TrainSet, mtry = 25, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 25
##
## OOB estimate of error rate: 31.91%
## Confusion matrix:
## 0 1 class.error
## 0 14 6 0.3000000
## 1 9 18 0.3333333importance(bag.model)## 0 1 MeanDecreaseAccuracy
## ENSG00000213606 3.4119081 4.18184753 4.7980805
## ENSG00000083857 -2.3325058 0.60009314 -1.0291516
## ENSG00000198074 4.1787200 3.13906948 4.4217293
## ENSG00000173391 2.1120207 3.28705124 3.6240647
## ENSG00000165912 3.2898648 0.94618109 2.8566631
## ENSG00000115009 2.8016300 2.54650587 3.4404161
## ENSG00000148773 -1.8597390 4.53521840 1.9759286
## ENSG00000129474 -1.1458335 -2.23387731 -2.2374472
## ENSG00000138448 -3.2578551 0.87179878 -1.6856044
## ENSG00000169429 5.8585756 2.77309059 5.8441789
## ENSG00000156535 -1.5287892 1.01351267 -0.3228246
## ENSG00000196177 -2.2274035 -0.74439925 -2.5920140
## ENSG00000060982 -1.0931995 0.98629542 0.2521196
## ENSG00000234964 0.5951718 1.62430958 1.5175364
## ENSG00000187498 0.2418235 -0.07158886 0.3329395
## ENSG00000231991 -0.3282091 -2.49841124 -1.8302299
## ENSG00000078098 1.7620838 5.96691677 4.7845545
## ENSG00000024526 8.9294135 4.14944926 8.5894947
## ENSG00000079931 -2.3194400 -0.10123742 -1.4942253
## ENSG00000137161 -0.4456163 -2.58316569 -2.4158183
## ENSG00000173744 -0.6035168 -2.53060566 -2.1518117
## ENSG00000118785 -1.5273488 -1.05196364 -1.4716860
## ENSG00000182718 1.7458746 0.69978147 2.2397654
## ENSG00000162366 2.4548491 -0.25691526 1.1587612
## ENSG00000168542 -1.2130002 1.32936662 0.1626110
## MeanDecreaseGini
## ENSG00000213606 1.6323679
## ENSG00000083857 0.2417435
## ENSG00000198074 2.6465439
## ENSG00000173391 1.0186995
## ENSG00000165912 1.4826157
## ENSG00000115009 0.5149525
## ENSG00000148773 1.4904747
## ENSG00000129474 0.3834768
## ENSG00000138448 0.2915831
## ENSG00000169429 2.8218948
## ENSG00000156535 0.1000760
## ENSG00000196177 1.2125862
## ENSG00000060982 1.0553121
## ENSG00000234964 0.8672632
## ENSG00000187498 0.2662054
## ENSG00000231991 0.1973183
## ENSG00000078098 1.1042246
## ENSG00000024526 3.1773363
## ENSG00000079931 0.1992244
## ENSG00000137161 0.5471829
## ENSG00000173744 0.2699355
## ENSG00000118785 0.1798585
## ENSG00000182718 0.1316691
## ENSG00000162366 0.3325649
## ENSG00000168542 0.3024648varImpPlot(bag.model)
Now use the model to predict the Testset outcomes
predict.bag <- predict(bag.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.bag)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 15 8
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 12.134, df = 1, p-value = 0.0004952As shown by the Chi2 and pval the model worked very well! but still lets see how random Forest will perform
Random Forest Method
lets look at several models using different numbers of features
model 1 = 6 features
rf1.model <- randomForest(defNASH ~ ., TrainSet, mtry= 6, importance= T )Vizualize the model results
rf1.model##
## Call:
## randomForest(formula = defNASH ~ ., data = TrainSet, mtry = 6, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 6
##
## OOB estimate of error rate: 31.91%
## Confusion matrix:
## 0 1 class.error
## 0 12 8 0.4000000
## 1 7 20 0.2592593importance(rf1.model)## 0 1 MeanDecreaseAccuracy
## ENSG00000213606 6.033528117 3.63088181 6.322180901
## ENSG00000083857 -1.395389109 3.54946467 2.219000413
## ENSG00000198074 3.493873614 0.51943614 2.488378991
## ENSG00000173391 4.366711351 3.44237183 5.024696947
## ENSG00000165912 4.089230617 3.88209930 5.576090497
## ENSG00000115009 4.158813127 2.51495612 3.901401862
## ENSG00000148773 2.858971900 3.51011216 4.559413647
## ENSG00000129474 0.288219511 -2.22444622 -1.533987712
## ENSG00000138448 -1.955460528 -0.64712944 -1.649364122
## ENSG00000169429 6.327500941 4.81684443 6.929213175
## ENSG00000156535 -0.740933082 -0.18103376 -0.579134325
## ENSG00000196177 1.096351441 -0.01542274 0.882641345
## ENSG00000060982 2.450122478 0.52021722 1.998026630
## ENSG00000234964 0.780289311 1.65341342 1.547601854
## ENSG00000187498 0.648058969 -0.72606174 -0.161619645
## ENSG00000231991 -1.912279489 -0.53835276 -1.601487199
## ENSG00000078098 2.101087639 2.44172071 2.794371550
## ENSG00000024526 7.346126737 4.04022548 7.028244304
## ENSG00000079931 0.020896294 -1.03350294 -0.939970024
## ENSG00000137161 1.366492818 -0.84708279 -0.002860047
## ENSG00000173744 -0.004035696 -0.45790973 -0.584667815
## ENSG00000118785 -1.871080210 -0.60661571 -1.644298369
## ENSG00000182718 -1.138400307 -0.98475508 -1.122012163
## ENSG00000162366 2.825345099 0.57980696 1.935222291
## ENSG00000168542 0.493033342 1.89711405 1.691999883
## MeanDecreaseGini
## ENSG00000213606 1.6573088
## ENSG00000083857 0.8170887
## ENSG00000198074 1.4602527
## ENSG00000173391 1.2516230
## ENSG00000165912 1.3269988
## ENSG00000115009 0.6767214
## ENSG00000148773 1.1160003
## ENSG00000129474 0.5303145
## ENSG00000138448 0.5527683
## ENSG00000169429 1.8662159
## ENSG00000156535 0.5393983
## ENSG00000196177 1.0735042
## ENSG00000060982 1.1887764
## ENSG00000234964 0.8780275
## ENSG00000187498 0.5062902
## ENSG00000231991 0.4586070
## ENSG00000078098 0.8978534
## ENSG00000024526 1.9888425
## ENSG00000079931 0.3988987
## ENSG00000137161 0.8531971
## ENSG00000173744 0.4114459
## ENSG00000118785 0.4641015
## ENSG00000182718 0.4591773
## ENSG00000162366 0.4144262
## ENSG00000168542 0.7522467varImpPlot(rf1.model)
plot(rf1.model)
Now use the model to predict the Testset outcomes
predict.rf <- predict(rf1.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.rf)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 16 7
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 14.141, df = 1, p-value = 0.0001696Model 1 about as good as bagging model
model 2 = 5 features
rf2.model <- randomForest(defNASH ~ ., TrainSet, mtry= 5, importance= T )Vizualize the model results
rf2.model##
## Call:
## randomForest(formula = defNASH ~ ., data = TrainSet, mtry = 5, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 5
##
## OOB estimate of error rate: 27.66%
## Confusion matrix:
## 0 1 class.error
## 0 14 6 0.3000000
## 1 7 20 0.2592593importance(rf2.model)## 0 1 MeanDecreaseAccuracy
## ENSG00000213606 3.11590105 2.1905363 3.35549384
## ENSG00000083857 -0.07336268 0.1828988 -0.01993989
## ENSG00000198074 3.61582091 3.2289963 4.21542342
## ENSG00000173391 3.32152677 5.0406765 5.74056630
## ENSG00000165912 3.14229713 4.9359925 5.10535489
## ENSG00000115009 3.61123014 0.8664017 2.91663340
## ENSG00000148773 2.68594003 4.1504435 4.72386266
## ENSG00000129474 -1.11461906 -2.5223644 -2.73387149
## ENSG00000138448 -0.31639910 -0.4755444 -0.38423434
## ENSG00000169429 2.78695598 5.1361860 5.96556448
## ENSG00000156535 -1.43628155 1.0483876 -0.15844836
## ENSG00000196177 -0.10065899 1.7153110 0.82560156
## ENSG00000060982 1.08215285 -0.2462770 0.19316944
## ENSG00000234964 -0.49064921 4.0325338 2.74701812
## ENSG00000187498 -1.55853652 -0.2875719 -1.32511070
## ENSG00000231991 -2.09252388 0.4581523 -1.00552164
## ENSG00000078098 1.24471389 3.6562970 3.60570017
## ENSG00000024526 8.17853046 6.0813947 8.76652826
## ENSG00000079931 -1.25359363 -0.3601234 -1.92414198
## ENSG00000137161 2.08443571 0.3551238 1.69167367
## ENSG00000173744 1.46716617 -2.1105797 -1.03452499
## ENSG00000118785 -1.23827851 -2.3968991 -2.31616698
## ENSG00000182718 0.41001708 -1.5302467 -1.16187978
## ENSG00000162366 3.41923389 1.1138149 2.81887172
## ENSG00000168542 -1.33768219 1.0256917 0.10532312
## MeanDecreaseGini
## ENSG00000213606 1.5302938
## ENSG00000083857 0.5286481
## ENSG00000198074 1.7938538
## ENSG00000173391 1.3737931
## ENSG00000165912 1.1408689
## ENSG00000115009 0.7168474
## ENSG00000148773 1.3763909
## ENSG00000129474 0.4536993
## ENSG00000138448 0.5821236
## ENSG00000169429 1.6329493
## ENSG00000156535 0.4618170
## ENSG00000196177 1.1301533
## ENSG00000060982 1.0935404
## ENSG00000234964 0.8781137
## ENSG00000187498 0.6542392
## ENSG00000231991 0.4395772
## ENSG00000078098 0.9800931
## ENSG00000024526 1.8198968
## ENSG00000079931 0.3637957
## ENSG00000137161 0.8006828
## ENSG00000173744 0.4964026
## ENSG00000118785 0.4656719
## ENSG00000182718 0.5275583
## ENSG00000162366 0.4430622
## ENSG00000168542 0.7642677varImpPlot(rf2.model)
plot(rf2.model)
Now use the model to predict the Testset outcomes
predict.rf <- predict(rf2.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.rf)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 16 7
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 14.141, df = 1, p-value = 0.00016965 features is better than 6
model 3 = 4 features
rf3.model <- randomForest(defNASH ~ ., TrainSet, mtry=4, importance= T )Vizualize the model results
rf3.model##
## Call:
## randomForest(formula = defNASH ~ ., data = TrainSet, mtry = 4, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 4
##
## OOB estimate of error rate: 27.66%
## Confusion matrix:
## 0 1 class.error
## 0 14 6 0.3000000
## 1 7 20 0.2592593importance(rf3.model)## 0 1 MeanDecreaseAccuracy
## ENSG00000213606 4.2974915 1.83014758 3.77537472
## ENSG00000083857 1.8890784 0.34535037 1.17407833
## ENSG00000198074 5.1506795 3.72160072 5.52440774
## ENSG00000173391 4.2609332 3.19306441 4.78521285
## ENSG00000165912 3.5673935 3.94470713 4.61079211
## ENSG00000115009 5.8363188 1.14035797 4.54711159
## ENSG00000148773 2.3969522 4.77005210 4.80713007
## ENSG00000129474 1.4371025 -1.56118961 -0.18016846
## ENSG00000138448 -0.2601430 0.77694059 0.01787410
## ENSG00000169429 2.6997836 3.49821260 3.87433263
## ENSG00000156535 -1.2424999 -0.56679119 -1.19787455
## ENSG00000196177 3.6609472 -0.58789651 1.93015582
## ENSG00000060982 0.1801544 1.69539674 1.19738180
## ENSG00000234964 3.0239648 4.66385734 5.13385036
## ENSG00000187498 -0.4164196 0.17194524 -0.44201979
## ENSG00000231991 0.6991853 -0.56542911 0.03814908
## ENSG00000078098 1.7177667 2.68874960 2.75976380
## ENSG00000024526 7.3342960 4.69089952 7.53021141
## ENSG00000079931 -1.8742482 -0.03452988 -1.33914984
## ENSG00000137161 1.1926255 0.11764821 0.52439947
## ENSG00000173744 1.2324301 -1.98221583 -0.88660102
## ENSG00000118785 -3.1240164 -1.01420094 -2.86607930
## ENSG00000182718 1.2612993 0.50157671 0.80728144
## ENSG00000162366 3.8771593 -0.28983760 1.85564627
## ENSG00000168542 0.7508060 2.75427484 2.37725744
## MeanDecreaseGini
## ENSG00000213606 1.4120397
## ENSG00000083857 0.7715356
## ENSG00000198074 1.5934935
## ENSG00000173391 1.2211124
## ENSG00000165912 1.1957401
## ENSG00000115009 0.8823206
## ENSG00000148773 1.1589112
## ENSG00000129474 0.5423547
## ENSG00000138448 0.5948034
## ENSG00000169429 1.4190480
## ENSG00000156535 0.6040489
## ENSG00000196177 0.9205694
## ENSG00000060982 1.3284664
## ENSG00000234964 0.9545475
## ENSG00000187498 0.6503239
## ENSG00000231991 0.4768369
## ENSG00000078098 0.8557878
## ENSG00000024526 1.8039866
## ENSG00000079931 0.5191374
## ENSG00000137161 0.6559942
## ENSG00000173744 0.4915720
## ENSG00000118785 0.5764491
## ENSG00000182718 0.4634314
## ENSG00000162366 0.5171184
## ENSG00000168542 0.8953922varImpPlot(rf3.model)
plot(rf3.model)
Now use the model to predict the Testset outcomes
predict.rf <- predict(rf3.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.rf)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 16 7
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 14.141, df = 1, p-value = 0.00016964 features same as 5 features
model 4 = 3 features
rf4.model <- randomForest(defNASH ~ ., TrainSet, mtry = 3, importance= T )Vizualize the model results
rf4.model##
## Call:
## randomForest(formula = defNASH ~ ., data = TrainSet, mtry = 3, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 3
##
## OOB estimate of error rate: 27.66%
## Confusion matrix:
## 0 1 class.error
## 0 15 5 0.2500000
## 1 8 19 0.2962963importance(rf4.model)## 0 1 MeanDecreaseAccuracy
## ENSG00000213606 4.5874651 3.4911454 5.402255222
## ENSG00000083857 -0.5308425 2.1364439 1.126240664
## ENSG00000198074 4.6093844 2.6089452 4.885664749
## ENSG00000173391 3.5886437 4.5230023 4.902255894
## ENSG00000165912 2.6974145 4.5386046 4.585519806
## ENSG00000115009 5.1925386 1.8311811 5.162500287
## ENSG00000148773 2.8347537 2.4720637 3.162338071
## ENSG00000129474 -0.7485664 -1.5641834 -1.629675228
## ENSG00000138448 -0.8996921 -0.9198803 -1.651649564
## ENSG00000169429 4.2684566 5.5945065 6.473263270
## ENSG00000156535 -2.8392588 -0.4765459 -2.251803002
## ENSG00000196177 -0.7681466 0.4804241 -0.086506717
## ENSG00000060982 1.5840341 1.6096992 2.014405995
## ENSG00000234964 1.8041471 2.3902237 2.716135738
## ENSG00000187498 -0.6073575 1.0790521 0.621871383
## ENSG00000231991 -1.4838539 0.3024591 -0.822746570
## ENSG00000078098 1.4694394 3.5066547 3.204142030
## ENSG00000024526 6.7702558 4.2838527 6.707241789
## ENSG00000079931 -0.3353972 -0.4639980 -0.581715679
## ENSG00000137161 3.0861283 1.1560082 2.390833474
## ENSG00000173744 0.2688846 -2.0076024 -1.477128669
## ENSG00000118785 -1.1332482 0.5776126 -0.334214542
## ENSG00000182718 0.6045072 -0.2429381 -0.004377996
## ENSG00000162366 2.9016499 1.6072362 3.112457261
## ENSG00000168542 1.0101595 1.0983792 1.232200955
## MeanDecreaseGini
## ENSG00000213606 1.2200044
## ENSG00000083857 0.8599981
## ENSG00000198074 1.6456633
## ENSG00000173391 1.1773349
## ENSG00000165912 1.1380842
## ENSG00000115009 0.7195974
## ENSG00000148773 1.1975523
## ENSG00000129474 0.6567509
## ENSG00000138448 0.5527933
## ENSG00000169429 1.5949500
## ENSG00000156535 0.5672682
## ENSG00000196177 1.1384986
## ENSG00000060982 1.0934111
## ENSG00000234964 0.9786290
## ENSG00000187498 0.6105460
## ENSG00000231991 0.5356567
## ENSG00000078098 0.8888001
## ENSG00000024526 1.5923593
## ENSG00000079931 0.5063392
## ENSG00000137161 0.8576097
## ENSG00000173744 0.4839495
## ENSG00000118785 0.6188079
## ENSG00000182718 0.4765763
## ENSG00000162366 0.5403537
## ENSG00000168542 0.7995296varImpPlot(rf4.model)
plot(rf4.model)
Now use the model to predict the Testset outcomes
predict.rf <- predict(rf4.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.rf)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 17 6
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 16.326, df = 1, p-value = 5.333e-053 features as good or better than 4 or 5
Now let’s compare the random models to a model based on the variable importance defined from the bagging approach using the top 3 features
model 5 top 4 features
rf5.model <- randomForest(defNASH ~ ENSG00000198074+ENSG00000196177+ENSG00000115009 +ENSG00000078098, TrainSet, importance= T )Vizualize the model results
rf5.model##
## Call:
## randomForest(formula = defNASH ~ ENSG00000198074 + ENSG00000196177 + ENSG00000115009 + ENSG00000078098, data = TrainSet, importance = T)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 27.66%
## Confusion matrix:
## 0 1 class.error
## 0 15 5 0.2500000
## 1 8 19 0.2962963importance(rf5.model)## 0 1 MeanDecreaseAccuracy MeanDecreaseGini
## ENSG00000198074 8.758810 3.7627259 8.286933 7.679080
## ENSG00000196177 4.810424 2.8079125 5.128385 5.541758
## ENSG00000115009 7.919137 4.1087794 8.652059 4.451713
## ENSG00000078098 6.140682 0.8775337 4.495579 4.776457varImpPlot(rf5.model)
plot(rf5.model)
Now use the model to predict the Testset outcomes
predict.rf <- predict(rf5.model, newdata = TestSet )
Model.Results <- table(observed = Testing.Outcome, predict = predict.rf)
Model.Results #confusion matrix## predict
## observed 0 1
## 0 16 7
## 1 4 21summary(Model.Results)## Number of cases in table: 48
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 14.141, df = 1, p-value = 0.0001696