Работа с пакетами CARET и BORUTA на языке R
Полина
Установка caret и визуальный анализ с featurePlot()
Установим и подключим пакет caret, необходимый для машинного обучения, просмотрим список доступных методов
names(getModelInfo())## [1] "ada" "AdaBag" "AdaBoost.M1"
## [4] "adaboost" "amdai" "ANFIS"
## [7] "avNNet" "awnb" "awtan"
## [10] "bag" "bagEarth" "bagEarthGCV"
## [13] "bagFDA" "bagFDAGCV" "bam"
## [16] "bartMachine" "bayesglm" "binda"
## [19] "blackboost" "blasso" "blassoAveraged"
## [22] "bridge" "brnn" "BstLm"
## [25] "bstSm" "bstTree" "C5.0"
## [28] "C5.0Cost" "C5.0Rules" "C5.0Tree"
## [31] "cforest" "chaid" "CSimca"
## [34] "ctree" "ctree2" "cubist"
## [37] "dda" "deepboost" "DENFIS"
## [40] "dnn" "dwdLinear" "dwdPoly"
## [43] "dwdRadial" "earth" "elm"
## [46] "enet" "evtree" "extraTrees"
## [49] "fda" "FH.GBML" "FIR.DM"
## [52] "foba" "FRBCS.CHI" "FRBCS.W"
## [55] "FS.HGD" "gam" "gamboost"
## [58] "gamLoess" "gamSpline" "gaussprLinear"
## [61] "gaussprPoly" "gaussprRadial" "gbm_h2o"
## [64] "gbm" "gcvEarth" "GFS.FR.MOGUL"
## [67] "GFS.LT.RS" "GFS.THRIFT" "glm.nb"
## [70] "glm" "glmboost" "glmnet_h2o"
## [73] "glmnet" "glmStepAIC" "gpls"
## [76] "hda" "hdda" "hdrda"
## [79] "HYFIS" "icr" "J48"
## [82] "JRip" "kernelpls" "kknn"
## [85] "knn" "krlsPoly" "krlsRadial"
## [88] "lars" "lars2" "lasso"
## [91] "lda" "lda2" "leapBackward"
## [94] "leapForward" "leapSeq" "Linda"
## [97] "lm" "lmStepAIC" "LMT"
## [100] "loclda" "logicBag" "LogitBoost"
## [103] "logreg" "lssvmLinear" "lssvmPoly"
## [106] "lssvmRadial" "lvq" "M5"
## [109] "M5Rules" "manb" "mda"
## [112] "Mlda" "mlp" "mlpKerasDecay"
## [115] "mlpKerasDecayCost" "mlpKerasDropout" "mlpKerasDropoutCost"
## [118] "mlpML" "mlpSGD" "mlpWeightDecay"
## [121] "mlpWeightDecayML" "monmlp" "msaenet"
## [124] "multinom" "mxnet" "mxnetAdam"
## [127] "naive_bayes" "nb" "nbDiscrete"
## [130] "nbSearch" "neuralnet" "nnet"
## [133] "nnls" "nodeHarvest" "null"
## [136] "OneR" "ordinalNet" "ordinalRF"
## [139] "ORFlog" "ORFpls" "ORFridge"
## [142] "ORFsvm" "ownn" "pam"
## [145] "parRF" "PART" "partDSA"
## [148] "pcaNNet" "pcr" "pda"
## [151] "pda2" "penalized" "PenalizedLDA"
## [154] "plr" "pls" "plsRglm"
## [157] "polr" "ppr" "pre"
## [160] "PRIM" "protoclass" "qda"
## [163] "QdaCov" "qrf" "qrnn"
## [166] "randomGLM" "ranger" "rbf"
## [169] "rbfDDA" "Rborist" "rda"
## [172] "regLogistic" "relaxo" "rf"
## [175] "rFerns" "RFlda" "rfRules"
## [178] "ridge" "rlda" "rlm"
## [181] "rmda" "rocc" "rotationForest"
## [184] "rotationForestCp" "rpart" "rpart1SE"
## [187] "rpart2" "rpartCost" "rpartScore"
## [190] "rqlasso" "rqnc" "RRF"
## [193] "RRFglobal" "rrlda" "RSimca"
## [196] "rvmLinear" "rvmPoly" "rvmRadial"
## [199] "SBC" "sda" "sdwd"
## [202] "simpls" "SLAVE" "slda"
## [205] "smda" "snn" "sparseLDA"
## [208] "spikeslab" "spls" "stepLDA"
## [211] "stepQDA" "superpc" "svmBoundrangeString"
## [214] "svmExpoString" "svmLinear" "svmLinear2"
## [217] "svmLinear3" "svmLinearWeights" "svmLinearWeights2"
## [220] "svmPoly" "svmRadial" "svmRadialCost"
## [223] "svmRadialSigma" "svmRadialWeights" "svmSpectrumString"
## [226] "tan" "tanSearch" "treebag"
## [229] "vbmpRadial" "vglmAdjCat" "vglmContRatio"
## [232] "vglmCumulative" "widekernelpls" "WM"
## [235] "wsrf" "xgbDART" "xgbLinear"
## [238] "xgbTree" "xyf"Сгенерируем искусственные данные для анализа
x <- matrix(rnorm(50*5),ncol=5)
y <- factor(rep(c("A", "B"), 25))
x## [,1] [,2] [,3] [,4] [,5]
## [1,] 0.62974610 -0.35720080 0.02506068 -0.625223988 1.623193914
## [2,] -0.27321789 0.88518083 1.36469115 0.372698060 1.678707118
## [3,] -1.47668504 -1.78672079 -0.49120546 0.583666993 -2.028311988
## [4,] -1.25377953 0.25563417 0.50470975 1.767352465 -0.771147800
## [5,] 1.17496705 -1.37603316 0.45079959 0.194090963 -0.540206766
## [6,] -0.86191037 -1.50043961 0.26111101 0.008013554 1.306285984
## [7,] -0.57620266 0.75927778 -1.33530571 -0.988155753 0.094697932
## [8,] -1.27967105 -0.55142810 -0.71855540 0.934718901 0.192682848
## [9,] 0.37760382 0.67698885 0.05267832 0.592179459 0.314734356
## [10,] -1.41700778 -0.27753303 1.42151284 1.176206744 -0.600321221
## [11,] -1.59218148 0.01923164 -0.56516601 0.560495829 -0.484961396
## [12,] -1.75448498 -0.27431020 0.27494570 0.227504729 -0.003682436
## [13,] 0.69065679 -0.53102061 0.18584605 -0.755741169 0.763286034
## [14,] 0.50631516 0.49232200 -0.08369772 0.412795040 -1.799830947
## [15,] 0.25969543 2.07336529 -0.68897046 0.236250104 1.329843624
## [16,] 1.08688416 1.13032301 0.13063584 -2.372401974 1.245402433
## [17,] -0.15822223 0.70647330 -1.18950039 -0.655723797 -2.269251096
## [18,] 0.45178352 -0.59870858 -0.02288185 -0.332044220 0.062775926
## [19,] 0.18472623 -0.59622598 0.68579465 -1.394711408 -0.401875203
## [20,] 0.63610363 -0.01043734 0.85523476 0.721390794 -0.121809505
## [21,] 0.16477993 0.73244956 1.27532005 0.360845142 0.726798823
## [22,] -0.97530771 1.30743916 0.30986153 0.366481768 -1.111512269
## [23,] -0.57834030 0.15149534 -0.26507641 0.425736401 -1.920625615
## [24,] -1.24965061 -3.35989024 1.94438871 -0.284513175 0.069522088
## [25,] 0.64598203 0.81308425 -0.69640165 1.105965567 1.499167529
## [26,] -0.10069538 -1.47311452 -1.42726643 -2.065333277 1.519253618
## [27,] -0.73481813 0.49108292 -0.32328446 -0.078188562 -0.699044213
## [28,] 0.52604108 -0.22056729 -0.90922691 -0.666118731 -0.510434282
## [29,] -2.08870793 -1.02878875 0.60771492 -0.180805363 -0.352192985
## [30,] 0.29764928 0.23236206 1.36918889 -0.458439389 -0.277223261
## [31,] 2.35009327 0.94894477 -1.44912205 -2.143188897 1.745763154
## [32,] -1.44558951 0.89839689 -1.28331975 -0.487762560 -0.546385687
## [33,] 1.43665693 1.90100018 1.34454274 -1.117920742 -2.799194457
## [34,] -0.44507245 0.62048417 0.34374815 0.859727864 0.194833197
## [35,] -0.07259785 -0.37762487 0.97040991 -0.562754045 -0.161007392
## [36,] 1.24941157 0.42594728 -1.52130461 -0.855872497 1.400017937
## [37,] -0.84795658 0.31556506 0.27893736 0.397127459 0.748783663
## [38,] 1.52676266 1.44766598 0.24526448 -0.030231802 -0.437247330
## [39,] 0.37135578 -0.24087657 0.50742558 0.158630789 1.236352619
## [40,] 1.02519481 -2.06457397 0.89473230 -0.768035338 1.482394690
## [41,] 1.15821918 0.53462585 0.72936078 -1.239626002 2.826997712
## [42,] -0.30764155 -1.14064087 0.76275687 -0.159624742 0.513553258
## [43,] 1.13516705 -0.24282885 1.11545077 2.052373804 0.536276856
## [44,] 1.21463914 0.53454116 -0.67551491 -1.697920670 -0.974734927
## [45,] 0.17470662 1.74875933 -0.05155020 -1.121087747 -1.603510965
## [46,] -0.25221700 -1.06458073 -1.03311000 -1.697191387 -0.783125686
## [47,] 0.08610853 -0.37409116 -1.14829264 1.121149318 0.105695065
## [48,] 1.50868578 -0.33996715 0.45772768 1.865556702 2.645931187
## [49,] -0.75665623 -2.21927937 1.55512783 -1.002530095 1.372320128
## [50,] 0.50355775 -1.27287897 0.69747679 -0.160929477 1.568882941y## [1] A B A B A B A B A B A B A B A B A B A B A B A B A B A B A B A B A B A B A B
## [39] A B A B A B A B A B A B
## Levels: A BВыполним разведочный графический анализ с помощью featurePlot()
featurePlot(x = x, y = y, plot = "density", auto.key = TRUE)
Оценка важности признаков через FSelector
Оценим информационный прирост признаков для набора данных iris
data(iris)
information.gain(Species ~ ., iris)## attr_importance
## Sepal.Length 0.4521286
## Sepal.Width 0.2672750
## Petal.Length 0.9402853
## Petal.Width 0.9554360Чем больше информационный прирост, тем важнее признак для разбиения классов
Преобразование непрерывной переменной в категориальную
Выполним преобразование 4 разными способами
interval - разбивает данные на равные интервалы, то есть длина интервала не зависит от количества значений, которые в него попадают
iris$PetalLength_interval <- discretize(iris$Petal.Length, method = "interval", breaks = 3)
table(iris$PetalLength_interval)##
## [1,2.97) [2.97,4.93) [4.93,6.9]
## 50 54 46frequency - разбивает данные так, чтобы в каждом интервале было одинаковое количество наблюдений, то есть интервалы будут иметь разную длину чтобы содержать равное количество данных
iris$PetalLength_frequency <- discretize(iris$Petal.Length, method = "frequency", breaks = 3)
table(iris$PetalLength_frequency)##
## [1,2.63) [2.63,4.9) [4.9,6.9]
## 50 49 51cluster - разбивает данные с использованием метода кластеризации, то есть метод ищет естественные группы данных и делит их на несколько категорий в зависимости от схожести значений
iris$PetalLength_cluster <- discretize(iris$Petal.Length, method = "cluster", breaks = 3)
table(iris$PetalLength_cluster)##
## [1,2.88) [2.88,4.96) [4.96,6.9]
## 50 54 46fixed - позволяет задавать собственные границы интервалам
iris$PetalLength_fixed <- discretize(iris$Petal.Length, method = "fixed", breaks = c(0, 2, 5, 7))
table(iris$PetalLength_fixed)##
## [0,2) [2,5) [5,7]
## 50 54 46Пакет Boruta и выбор признаков
Удалим строки с нулевыми значениями и выполним функцию выборки важных признаков, по результатам её выполнения построим boxplot
data("Ozone")
ozone_data <- na.omit(Ozone)
boruta_output <- Boruta(V4 ~ ., data = ozone_data, doTrace = 1)## After 11 iterations, +1.4 secs:## confirmed 9 attributes: V1, V10, V11, V12, V13 and 4 more;## rejected 1 attribute: V3;## still have 2 attributes left.## After 18 iterations, +1.9 secs:## rejected 1 attribute: V2;## still have 1 attribute left.## After 27 iterations, +2.6 secs:## rejected 1 attribute: V6;## no more attributes left.plot(boruta_output, las = 2, cex.axis = 0.7)