0
我想通过使用线性SVM模型与RBF内核模型测试我的数据来测试我的数据是否可线性分离，以查看两个使用F分数之间哪个分数更好。致命错误与插入符号和RStudio
我使用了脱字符号包，并且我设置了两个模型，fit.SVMKernel使用'rbf'方法，而fit.SVML使用'svmLinear'。
我可以运行我的脚本的整个长度直到线性模型，并执行没有任何问题，但只要我运行内核模型，我得到一个致命的错误消息，并必须重新启动会话，请参阅附图下面。
任何人都可以提供一个建议，为什么R每次运行内核代码段时崩溃？
这是我的data和我的code的可重复版本给任何人提供两个hoots！
我已加载插入符包按以下
> sessionInfo()
R version 3.3.2 (2016-10-31)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)
locale:
[1] LC_COLLATE=English_Ireland.1252 LC_CTYPE=English_Ireland.1252 LC_MONETARY=English_Ireland.1252 LC_NUMERIC=C
[5] LC_TIME=English_Ireland.1252
attached base packages:
[1] parallel stats graphics grDevices utils datasets methods base
other attached packages:
[1] kernlab_0.9-25 doMC_1.3.4 iterators_1.0.8 foreach_1.4.3 Amelia_1.7.4 Rcpp_0.12.9 dplyr_0.5.0 pROC_1.9.1 klaR_0.6-12
[10] MASS_7.3-45 caret_6.0-73 ggplot2_2.2.1 lattice_0.20-34
loaded via a namespace (and not attached):
[1] compiler_3.3.2 nloptr_1.0.4 plyr_1.8.4 class_7.3-14 tools_3.3.2 digest_0.6.12 lme4_1.1-12 tibble_1.2
[9] nlme_3.1-128 gtable_0.2.0 mgcv_1.8-15 Matrix_1.2-7.1 DBI_0.5-1 SparseM_1.74 e1071_1.6-8 stringr_1.2.0
[17] MatrixModels_0.4-1 stats4_3.3.2 combinat_0.0-8 grid_3.3.2 nnet_7.3-12 R6_2.2.0 foreign_0.8-67 minqa_1.2.4
[25] reshape2_1.4.2 car_2.1-4 magrittr_1.5 scales_0.4.1 codetools_0.2-15 ModelMetrics_1.1.0 splines_3.3.2 assertthat_0.1
[33] pbkrtest_0.4-6 colorspace_1.3-2 labeling_0.3 quantreg_5.29 stringi_1.1.2 lazyeval_0.2.0 munsell_0.4.3
结果变量我想在分类（training.data.raw $超调）的类型的因素：
> str(training.data.raw)
'data.frame': 2846 obs. of 19 variables:
$ Total.Tx.Height : num 31.2 31.2 31.2 31.2 31.2 ...
$ Antenna.Tilt : int 0 0 0 0 0 0 4 4 2 2 ...
$ Antenna.Gain : num 15.9 15.9 15.9 18.2 18.2 18.2 15.9 15.9 18.8 18.8 ...
$ Ant.Vert.Beamwidth: num 10 10 10 4.4 4.4 4.4 9.6 9.6 4.3 4.3 ...
$ RTWP : num -106 -104 -105 -105 -105 ...
$ Voice.Drops : int 1 12 1 0 1 5 1 18 4 3 ...
$ Range : num 11.33 5.14 5.14 11.33 3.88 ...
$ Max.Distance : num 12.43 6.24 6.24 12.43 4.98 ...
$ Environment : Factor w/ 3 levels "Rural","Suburban",..: 1 1 1 1 1 1 1 1 1 1 ...
$ Rural : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 2 2 2 2 ...
$ Suburban : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ Urban : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ Overshooting : Factor w/ 2 levels "0","1": 1 1 2 2 2 2 2 2 2 1 ...
$ OShooting : Factor w/ 2 levels "Not.Overshooting",..: 1 1 2 2 2 2 2 2 2 1 ...
$ HSUPA.Throughput : num 164 223 232 241 264 ...
$ Max.HSDPA.Users : int 10 16 9 5 8 7 14 31 8 12 ...
$ HS.DSCH.throughput: num 1975 2346 2995 3696 3894 ...
$ Max.HSUPA.Users : int 13 25 11 5 13 9 15 33 8 13 ...
$ Avg.CQI : num 16.2 18 19.4 19.7 21.8 ...
以下是我用于两种型号的训练控制的代码：
metric <- "Accuracy"
train_Control <- trainControl(method = "repeatedCV",
number = 10,
repeats = 3),
classProbs = T)
这里有两种模式，我创建fit.SVMKernel和fit.SVMLinear
set.seed(123)
fit.SVML <- train(Overshooting ~ Total.Tx.Height +
Antenna.Tilt +
Antenna.Gain +
Ant.Vert.Beamwidth +
RTWP +
Voice.Drops +
Range +
Max.Distance +
Rural +
Suburban +
Urban +
HSUPA.Throughput +
Max.HSDPA.Users +
HS.DSCH.throughput +
Max.HSUPA.Users +
Avg.CQI,
data = training.data.raw,
method = 'svmLinear',
preProcess = c('center','scale'),
trControl=train_Control,
tuneLength=5,
metric = metric)
set.seed(123)
fit.SVMKernel <- train(Overshooting ~ Total.Tx.Height +
Antenna.Tilt +
Antenna.Gain +
Ant.Vert.Beamwidth +
RTWP +
Voice.Drops +
Range +
Max.Distance +
Rural +
Suburban +
Urban +
HSUPA.Throughput +
Max.HSDPA.Users +
HS.DSCH.throughput +
Max.HSUPA.Users +
Avg.CQI,
data = training.data.raw,
method = 'rbf',
preProcess = c('center','scale'),
trControl=train_Control,
tuneLength=5,
metric = metric)#,
#summaryFunction=twoClassSummary)