Caso de Estudo Mnist
Luís
09/05/2020
1.Exportação de Dados
library(dslabs)
library(matrixStats)
library(tidyverse)## -- Attaching packages ---------------------------------------------------------------------------------------------------------------- tidyverse 1.3.0 --## v ggplot2 3.2.1 v purrr 0.3.3
## v tibble 2.1.3 v dplyr 0.8.3
## v tidyr 1.0.0 v stringr 1.4.0
## v readr 1.3.1 v forcats 0.4.0## -- Conflicts ------------------------------------------------------------------------------------------------------------------- tidyverse_conflicts() --
## x dplyr::count() masks matrixStats::count()
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()library(dslabs)
library(GGally)## Registered S3 method overwritten by 'GGally':
## method from
## +.gg ggplot2##
## Attaching package: 'GGally'## The following object is masked from 'package:dplyr':
##
## nasalibrary(caret)## Loading required package: lattice##
## Attaching package: 'caret'## The following object is masked from 'package:purrr':
##
## liftlibrary(purrr)
library(class)
library(Rborist)## Rborist 0.2-3## Type RboristNews() to see new features/changes/bug fixes.library(rafalib)
library(tidyverse)
mnist <- read_mnist()
names(mnist)## [1] "train" "test"dim(mnist$train$images)## [1] 60000 784class(mnist$train$labels)## [1] "integer"table(mnist$train$labels)##
## 0 1 2 3 4 5 6 7 8 9
## 5923 6742 5958 6131 5842 5421 5918 6265 5851 5949# sample 10k rows from training set, 1k rows from test set
set.seed(123)
index <- sample(nrow(mnist$train$images), 10000)
x <- mnist$train$images[index,]
y <- factor(mnist$train$labels[index])
index <- sample(nrow(mnist$test$images), 1000)
#note that the line above is the corrected code - code in video at 0:52 is incorrect
x_test <- mnist$test$images[index,]
y_test <- factor(mnist$test$labels[index])- Pré-Processamento
sds <- colSds(x)
qplot(sds, bins = 256, color = I('black'))
col_index <- setdiff(1:ncol(x), x)
length(col_index)## [1] 529nzv <- nearZeroVar(x)
image(matrix(1:784 %in% nzv, 28, 28))
image(matrix(1:784 %in% x, 28, 28))
col_index <- setdiff(1:ncol(x), nzv)
length(col_index)## [1] 249- KNN e Random Forest
colnames(x) <- 1:ncol(mnist$train$images)
colnames(x_test) <- colnames(x)
#control <- trainControl(method = "cv", number = 10, p = .9)
#train_knn <- train(x[,col_index], y,
# method = "knn",
# tuneGrid = data.frame(k = c(1,3,5,7)),
# trControl = control)
#ggplot(train_knn)
n <- 1000
b <- 2
index <- sample(nrow(x), n)
control <- trainControl(method = "cv", number = b, p = .9)
train_knn <- train(x[index ,col_index], y[index],
method = "knn",
tuneGrid = data.frame(k = c(3,5,7)),
trControl = control)
fit_knn <- knn3(x[ ,col_index], y, k = 3)
y_hat_knn <- predict(fit_knn,
x_test[, col_index],
type="class")
cm <- confusionMatrix(y_hat_knn, factor(y_test))
cm## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2 3 4 5 6 7 8 9
## 0 106 0 0 0 0 1 3 0 0 2
## 1 0 111 1 0 0 1 2 1 1 0
## 2 0 1 89 0 0 0 0 0 1 1
## 3 0 0 0 101 0 4 0 0 4 0
## 4 0 0 0 0 82 0 0 0 1 4
## 5 0 0 0 5 0 82 1 0 0 0
## 6 1 0 2 1 0 1 101 0 2 0
## 7 1 0 4 1 0 0 0 96 0 0
## 8 0 0 0 0 1 4 0 0 88 0
## 9 0 0 0 0 2 1 0 2 0 87
##
## Overall Statistics
##
## Accuracy : 0.943
## 95% CI : (0.9268, 0.9565)
## No Information Rate : 0.112
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9366
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 0.9815 0.9911 0.9271 0.9352 0.9647 0.8723
## Specificity 0.9933 0.9932 0.9967 0.9910 0.9945 0.9934
## Pos Pred Value 0.9464 0.9487 0.9674 0.9266 0.9425 0.9318
## Neg Pred Value 0.9977 0.9989 0.9923 0.9921 0.9967 0.9868
## Prevalence 0.1080 0.1120 0.0960 0.1080 0.0850 0.0940
## Detection Rate 0.1060 0.1110 0.0890 0.1010 0.0820 0.0820
## Detection Prevalence 0.1120 0.1170 0.0920 0.1090 0.0870 0.0880
## Balanced Accuracy 0.9874 0.9922 0.9619 0.9631 0.9796 0.9329
## Class: 6 Class: 7 Class: 8 Class: 9
## Sensitivity 0.9439 0.9697 0.9072 0.9255
## Specificity 0.9922 0.9933 0.9945 0.9945
## Pos Pred Value 0.9352 0.9412 0.9462 0.9457
## Neg Pred Value 0.9933 0.9967 0.9901 0.9923
## Prevalence 0.1070 0.0990 0.0970 0.0940
## Detection Rate 0.1010 0.0960 0.0880 0.0870
## Detection Prevalence 0.1080 0.1020 0.0930 0.0920
## Balanced Accuracy 0.9680 0.9815 0.9508 0.9600control <- trainControl(method="cv", number = 5, p = 0.8)
grid <- expand.grid(minNode = c(1,5) , predFixed = c(10, 15, 25, 35, 50))
train_rf <- train(x[, col_index], y,
method = "Rborist",
nTree = 50,
trControl = control,
tuneGrid = grid,
nSamp = 5000)
ggplot(train_rf)
train_rf$bestTune## predFixed minNode
## 2 15 1fit_rf <- Rborist(x[, col_index], y,
nTree = 1000,
minNode = train_rf$bestTune$minNode,
predFixed = train_rf$bestTune$predFixed)
y_hat_rf <- factor(levels(y)[predict(fit_rf, x_test[ ,col_index])$yPred])
cm <- confusionMatrix(y_hat_rf, y_test)
cm## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2 3 4 5 6 7 8 9
## 0 108 0 0 0 0 3 0 0 0 1
## 1 0 110 0 0 0 1 1 0 0 0
## 2 0 2 93 0 0 1 0 3 0 0
## 3 0 0 0 105 0 3 1 0 0 1
## 4 0 0 0 0 82 0 2 1 1 3
## 5 0 0 0 2 0 83 1 0 1 2
## 6 0 0 1 1 0 0 101 0 1 0
## 7 0 0 2 0 0 1 0 93 0 0
## 8 0 0 0 0 1 2 1 0 94 0
## 9 0 0 0 0 2 0 0 2 0 87
##
## Overall Statistics
##
## Accuracy : 0.956
## 95% CI : (0.9414, 0.9679)
## No Information Rate : 0.112
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9511
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 1.0000 0.9821 0.9688 0.9722 0.9647 0.8830
## Specificity 0.9955 0.9977 0.9934 0.9944 0.9923 0.9934
## Pos Pred Value 0.9643 0.9821 0.9394 0.9545 0.9213 0.9326
## Neg Pred Value 1.0000 0.9977 0.9967 0.9966 0.9967 0.9879
## Prevalence 0.1080 0.1120 0.0960 0.1080 0.0850 0.0940
## Detection Rate 0.1080 0.1100 0.0930 0.1050 0.0820 0.0830
## Detection Prevalence 0.1120 0.1120 0.0990 0.1100 0.0890 0.0890
## Balanced Accuracy 0.9978 0.9899 0.9811 0.9833 0.9785 0.9382
## Class: 6 Class: 7 Class: 8 Class: 9
## Sensitivity 0.9439 0.9394 0.9691 0.9255
## Specificity 0.9966 0.9967 0.9956 0.9956
## Pos Pred Value 0.9712 0.9687 0.9592 0.9560
## Neg Pred Value 0.9933 0.9934 0.9967 0.9923
## Prevalence 0.1070 0.0990 0.0970 0.0940
## Detection Rate 0.1010 0.0930 0.0940 0.0870
## Detection Prevalence 0.1040 0.0960 0.0980 0.0910
## Balanced Accuracy 0.9703 0.9680 0.9823 0.9606rafalib::mypar(3,4)
for(i in 1:12){
image(matrix(x_test[i,], 28, 28)[, 28:1],
main = paste("Our prediction:", y_hat_rf[i]),
xaxt="n", yaxt="n")
}
## Variaveis mais Importantes
library(randomForest)## randomForest 4.6-14## Type rfNews() to see new features/changes/bug fixes.##
## Attaching package: 'randomForest'## The following object is masked from 'package:dplyr':
##
## combine## The following object is masked from 'package:ggplot2':
##
## marginx <- mnist$train$images[index,]
y <- factor(mnist$train$labels[index])
rf <- randomForest(x, y, ntree = 50)
imp <- importance(rf)
imp## MeanDecreaseGini
## 1 0.00000000
## 2 0.00000000
## 3 0.00000000
## 4 0.00000000
## 5 0.00000000
## 6 0.00000000
## 7 0.00000000
## 8 0.00000000
## 9 0.00000000
## 10 0.00000000
## 11 0.00000000
## 12 0.00000000
## 13 0.00000000
## 14 0.00000000
## 15 0.00000000
## 16 0.00000000
## 17 0.00000000
## 18 0.00000000
## 19 0.00000000
## 20 0.00000000
## 21 0.00000000
## 22 0.00000000
## 23 0.00000000
## 24 0.00000000
## 25 0.00000000
## 26 0.00000000
## 27 0.00000000
## 28 0.00000000
## 29 0.00000000
## 30 0.00000000
## 31 0.00000000
## 32 0.00000000
## 33 0.00000000
## 34 0.00000000
## 35 0.00000000
## 36 0.00000000
## 37 0.00000000
## 38 0.00000000
## 39 0.00000000
## 40 0.00000000
## 41 0.00000000
## 42 0.00000000
## 43 0.00000000
## 44 0.03516667
## 45 0.00000000
## 46 0.00000000
## 47 0.00000000
## 48 0.00000000
## 49 0.00000000
## 50 0.00000000
## 51 0.00000000
## 52 0.00000000
## 53 0.00000000
## 54 0.00000000
## 55 0.00000000
## 56 0.00000000
## 57 0.00000000
## 58 0.00000000
## 59 0.00000000
## 60 0.00000000
## 61 0.00000000
## 62 0.00000000
## 63 0.02000000
## 64 0.00000000
## 65 0.00000000
## 66 0.00000000
## 67 0.04636364
## 68 0.04000000
## 69 0.08689474
## 70 0.36721179
## 71 0.05200000
## 72 0.09434792
## 73 0.00000000
## 74 0.18555547
## 75 0.00000000
## 76 0.12999574
## 77 0.00000000
## 78 0.00000000
## 79 0.00000000
## 80 0.00000000
## 81 0.00000000
## 82 0.00000000
## 83 0.00000000
## 84 0.00000000
## 85 0.00000000
## 86 0.00000000
## 87 0.00000000
## 88 0.00000000
## 89 0.00000000
## 90 0.03733333
## 91 0.00000000
## 92 0.00000000
## 93 0.00000000
## 94 0.09568627
## 95 0.00000000
## 96 0.13506603
## 97 0.30127083
## 98 0.28552413
## 99 0.73202957
## 100 1.08493503
## 101 4.68703988
## 102 5.04271550
## 103 1.57166777
## 104 1.57234645
## 105 0.08400000
## 106 0.15950034
## 107 0.00000000
## 108 0.00000000
## 109 0.00000000
## 110 0.00000000
## 111 0.00000000
## 112 0.00000000
## 113 0.00000000
## 114 0.00000000
## 115 0.00000000
## 116 0.00000000
## 117 0.00000000
## 118 0.06163478
## 119 0.00000000
## 120 0.05851852
## 121 0.22669765
## 122 0.82722330
## 123 0.27007269
## 124 0.86211226
## 125 1.52033476
## 126 0.96086099
## 127 1.50438791
## 128 3.54764249
## 129 1.01390667
## 130 1.43391587
## 131 0.56584532
## 132 0.54811401
## 133 0.48051006
## 134 0.22830974
## 135 0.13119349
## 136 0.07692308
## 137 0.00000000
## 138 0.00000000
## 139 0.00000000
## 140 0.00000000
## 141 0.00000000
## 142 0.00000000
## 143 0.00000000
## 144 0.00000000
## 145 0.04600000
## 146 0.00000000
## 147 0.07400000
## 148 0.11866667
## 149 0.92668926
## 150 1.40555513
## 151 4.68572705
## 152 4.17026173
## 153 5.87704055
## 154 1.91873099
## 155 4.40723963
## 156 5.71669443
## 157 4.35657287
## 158 1.38923402
## 159 2.21977137
## 160 1.40883234
## 161 1.03124587
## 162 0.57810047
## 163 0.47184358
## 164 0.03780952
## 165 0.30549994
## 166 0.07373333
## 167 0.00000000
## 168 0.00000000
## 169 0.00000000
## 170 0.00000000
## 171 0.00000000
## 172 0.00000000
## 173 0.09370019
## 174 0.00000000
## 175 0.27557843
## 176 0.52792605
## 177 1.40932456
## 178 2.79206099
## 179 1.77499637
## 180 1.24747327
## 181 3.31881264
## 182 3.07029102
## 183 2.82974892
## 184 4.37295461
## 185 2.64050607
## 186 2.79952867
## 187 1.59238504
## 188 1.18604264
## 189 1.49483845
## 190 0.44617275
## 191 0.93783638
## 192 0.53405921
## 193 0.39931400
## 194 0.00000000
## 195 0.00000000
## 196 0.00000000
## 197 0.00000000
## 198 0.00000000
## 199 0.00000000
## 200 0.00000000
## 201 0.02000000
## 202 0.14943124
## 203 0.33511584
## 204 0.53502464
## 205 1.79593215
## 206 2.83933854
## 207 2.24164647
## 208 3.91093248
## 209 2.05235690
## 210 2.71847911
## 211 6.38065178
## 212 6.23179650
## 213 4.68900448
## 214 1.73924526
## 215 1.56738867
## 216 1.58952455
## 217 1.32875984
## 218 1.08206621
## 219 1.55005787
## 220 0.88038331
## 221 0.68676631
## 222 0.13601504
## 223 0.03333333
## 224 0.00000000
## 225 0.00000000
## 226 0.00000000
## 227 0.00000000
## 228 0.00000000
## 229 0.15600000
## 230 0.42974259
## 231 0.35458177
## 232 1.13432504
## 233 1.03361790
## 234 0.85646594
## 235 2.42345161
## 236 1.78527110
## 237 2.28375888
## 238 3.04775075
## 239 1.87318410
## 240 5.83392466
## 241 2.74398802
## 242 3.00501290
## 243 2.53730745
## 244 2.07290226
## 245 1.48601327
## 246 1.42769332
## 247 1.35162673
## 248 0.35846852
## 249 1.15078033
## 250 0.15242910
## 251 0.00000000
## 252 0.00000000
## 253 0.00000000
## 254 0.00000000
## 255 0.10892273
## 256 0.00000000
## 257 0.00000000
## 258 0.11460423
## 259 0.67189363
## 260 1.05929168
## 261 0.87284662
## 262 1.47123503
## 263 2.57743127
## 264 4.58818331
## 265 3.03614642
## 266 2.70862633
## 267 2.74009087
## 268 3.01405686
## 269 1.51488979
## 270 2.55274141
## 271 3.28748755
## 272 1.69413259
## 273 3.05533853
## 274 0.76321495
## 275 1.22615232
## 276 0.52747514
## 277 0.20914883
## 278 0.10381702
## 279 0.00000000
## 280 0.00000000
## 281 0.00000000
## 282 0.00000000
## 283 0.00000000
## 284 0.03544118
## 285 0.09444884
## 286 0.47771035
## 287 0.70334009
## 288 0.87812128
## 289 0.40111496
## 290 2.92116840
## 291 3.66874000
## 292 4.01775784
## 293 2.43258606
## 294 1.71425016
## 295 1.33642211
## 296 3.64906830
## 297 3.07775929
## 298 2.98491008
## 299 2.39849334
## 300 3.15774539
## 301 3.41660471
## 302 0.79485547
## 303 0.69870281
## 304 0.12891228
## 305 0.03911111
## 306 0.00000000
## 307 0.00000000
## 308 0.00000000
## 309 0.00000000
## 310 0.00000000
## 311 0.03709756
## 312 0.00000000
## 313 0.00000000
## 314 0.23500204
## 315 0.34771543
## 316 0.37812121
## 317 1.49731198
## 318 3.39817158
## 319 3.21356729
## 320 4.30391475
## 321 2.79338704
## 322 3.23442598
## 323 4.53620675
## 324 6.23768947
## 325 3.09680293
## 326 3.81904880
## 327 4.50140604
## 328 3.10128683
## 329 1.92431416
## 330 1.35260449
## 331 1.93857348
## 332 0.05000000
## 333 0.05538462
## 334 0.00000000
## 335 0.00000000
## 336 0.00000000
## 337 0.00000000
## 338 0.00000000
## 339 0.00000000
## 340 0.00000000
## 341 0.06733333
## 342 0.08484848
## 343 0.36767341
## 344 1.20439248
## 345 3.13319339
## 346 3.06492213
## 347 4.98804289
## 348 6.07968895
## 349 4.01331000
## 350 4.68493310
## 351 9.46559718
## 352 1.93826110
## 353 3.69963454
## 354 3.48648385
## 355 2.53188935
## 356 1.54225652
## 357 0.71324339
## 358 2.22883633
## 359 5.48484248
## 360 1.01508920
## 361 0.00000000
## 362 0.00000000
## 363 0.00000000
## 364 0.00000000
## 365 0.00000000
## 366 0.00000000
## 367 0.00000000
## 368 0.00000000
## 369 0.02000000
## 370 0.29563294
## 371 0.82341385
## 372 1.31551836
## 373 2.60617109
## 374 8.72354804
## 375 4.39793196
## 376 4.05508066
## 377 3.66204560
## 378 7.14313149
## 379 11.47782692
## 380 4.02405923
## 381 6.65865191
## 382 5.35600229
## 383 4.97977124
## 384 1.33112215
## 385 1.41951181
## 386 2.24618133
## 387 1.97624048
## 388 0.09466667
## 389 0.28902944
## 390 0.00000000
## 391 0.00000000
## 392 0.00000000
## 393 0.00000000
## 394 0.00000000
## 395 0.00000000
## 396 0.07704954
## 397 0.05777778
## 398 0.40708408
## 399 1.02550425
## 400 2.28213028
## 401 3.46655168
## 402 3.27674110
## 403 5.86298444
## 404 7.20104787
## 405 2.64313252
## 406 3.69009971
## 407 7.93163228
## 408 4.77081852
## 409 4.13594681
## 410 5.66789399
## 411 4.13182152
## 412 0.75392899
## 413 0.84953320
## 414 1.21134441
## 415 1.93291884
## 416 0.09549020
## 417 0.13028571
## 418 0.00000000
## 419 0.00000000
## 420 0.00000000
## 421 0.00000000
## 422 0.00000000
## 423 0.00000000
## 424 0.00000000
## 425 0.04000000
## 426 0.30841684
## 427 0.81554285
## 428 3.60507602
## 429 5.50947189
## 430 3.60975038
## 431 4.96652344
## 432 4.88709906
## 433 2.93491528
## 434 5.46926986
## 435 7.06636675
## 436 5.08719654
## 437 2.90371908
## 438 7.27514142
## 439 2.99415287
## 440 1.54610067
## 441 1.40235844
## 442 2.83833042
## 443 1.28776026
## 444 0.19133568
## 445 0.14320938
## 446 0.09331263
## 447 0.00000000
## 448 0.00000000
## 449 0.00000000
## 450 0.00000000
## 451 0.00000000
## 452 0.03733333
## 453 0.03733333
## 454 0.32777839
## 455 0.96981953
## 456 3.02911828
## 457 3.22458532
## 458 2.51985596
## 459 3.67894390
## 460 2.62268307
## 461 3.31313400
## 462 8.28346874
## 463 4.62456791
## 464 3.91809077
## 465 4.92617301
## 466 4.32000975
## 467 4.45803263
## 468 0.99933785
## 469 1.84532901
## 470 2.06830432
## 471 0.69833335
## 472 0.26209607
## 473 0.00000000
## 474 0.06933333
## 475 0.00000000
## 476 0.00000000
## 477 0.00000000
## 478 0.00000000
## 479 0.00000000
## 480 0.03000000
## 481 0.11140741
## 482 1.02771869
## 483 1.97080441
## 484 2.95609582
## 485 2.51083357
## 486 3.46723910
## 487 3.68246577
## 488 2.90238826
## 489 5.78384945
## 490 5.77159785
## 491 3.80868416
## 492 1.71665004
## 493 3.75454800
## 494 4.07300152
## 495 2.65116795
## 496 1.94406880
## 497 1.75076673
## 498 1.91859822
## 499 0.96021324
## 500 0.40140175
## 501 0.25056246
## 502 0.07363636
## 503 0.00000000
## 504 0.00000000
## 505 0.00000000
## 506 0.02000000
## 507 0.00000000
## 508 0.00000000
## 509 0.32431244
## 510 0.13021068
## 511 3.36154083
## 512 3.35951337
## 513 2.23100073
## 514 3.19548488
## 515 5.85639361
## 516 4.15502382
## 517 1.99351557
## 518 3.75605123
## 519 2.18299992
## 520 1.83143505
## 521 1.35589572
## 522 1.92326430
## 523 1.96546328
## 524 2.25231128
## 525 1.15158083
## 526 0.91700287
## 527 0.89076084
## 528 0.50975551
## 529 0.37852677
## 530 0.00000000
## 531 0.00000000
## 532 0.00000000
## 533 0.00000000
## 534 0.00000000
## 535 0.00000000
## 536 0.05804356
## 537 0.77476242
## 538 0.67624571
## 539 3.33551360
## 540 4.16432974
## 541 2.21982230
## 542 2.19345087
## 543 4.57817152
## 544 6.95054479
## 545 2.92951925
## 546 1.85221842
## 547 1.17263405
## 548 1.49532249
## 549 2.02780060
## 550 2.99963359
## 551 2.46476172
## 552 1.28620386
## 553 0.87180205
## 554 1.23930996
## 555 0.60914594
## 556 0.31514449
## 557 0.19743834
## 558 0.16860450
## 559 0.00000000
## 560 0.00000000
## 561 0.00000000
## 562 0.00000000
## 563 0.00000000
## 564 0.09333333
## 565 0.05200000
## 566 0.49910464
## 567 2.57215427
## 568 2.36390248
## 569 1.82455746
## 570 1.59861240
## 571 3.27335049
## 572 2.53209265
## 573 2.57830880
## 574 2.05509207
## 575 1.60418257
## 576 1.90174687
## 577 1.80839016
## 578 2.10968678
## 579 2.20883703
## 580 0.82420996
## 581 0.85477704
## 582 0.38647854
## 583 0.37883033
## 584 0.22213566
## 585 0.00000000
## 586 0.00000000
## 587 0.00000000
## 588 0.00000000
## 589 0.00000000
## 590 0.00000000
## 591 0.00000000
## 592 0.05100000
## 593 0.42675075
## 594 0.23786667
## 595 0.92749849
## 596 2.71354684
## 597 4.85195543
## 598 2.22579617
## 599 2.95441913
## 600 1.75202828
## 601 1.72084819
## 602 1.74954019
## 603 1.00692574
## 604 1.25952725
## 605 1.58921090
## 606 1.63939731
## 607 1.97333414
## 608 1.62371665
## 609 0.49585811
## 610 0.69558859
## 611 0.46934290
## 612 0.00000000
## 613 0.00000000
## 614 0.00000000
## 615 0.00000000
## 616 0.00000000
## 617 0.00000000
## 618 0.00000000
## 619 0.00000000
## 620 0.00000000
## 621 0.03733333
## 622 0.30914575
## 623 0.68692214
## 624 0.88368123
## 625 2.62290773
## 626 3.55930591
## 627 4.44395957
## 628 1.67945553
## 629 2.43707112
## 630 2.16416834
## 631 0.85225581
## 632 1.76399346
## 633 1.40422280
## 634 1.25624628
## 635 0.64721428
## 636 0.59034628
## 637 0.34288740
## 638 0.40944949
## 639 0.17289779
## 640 0.10228600
## 641 0.05714286
## 642 0.00000000
## 643 0.00000000
## 644 0.00000000
## 645 0.00000000
## 646 0.00000000
## 647 0.00000000
## 648 0.00000000
## 649 0.02000000
## 650 0.07333333
## 651 0.11492052
## 652 0.13397358
## 653 1.36191863
## 654 1.58652064
## 655 2.26954182
## 656 2.48726366
## 657 2.51786456
## 658 2.91050569
## 659 2.25122232
## 660 1.09642982
## 661 0.73781301
## 662 1.06052889
## 663 0.53397481
## 664 0.63035287
## 665 0.09881212
## 666 0.11050549
## 667 0.03754032
## 668 0.00000000
## 669 0.00000000
## 670 0.00000000
## 671 0.00000000
## 672 0.00000000
## 673 0.00000000
## 674 0.00000000
## 675 0.00000000
## 676 0.00000000
## 677 0.00000000
## 678 0.02000000
## 679 0.07297571
## 680 0.24664706
## 681 0.57726440
## 682 0.42999030
## 683 1.25945514
## 684 0.94466864
## 685 0.76116948
## 686 0.70789956
## 687 1.00959426
## 688 0.44306012
## 689 0.71795132
## 690 0.60214548
## 691 0.30874703
## 692 0.20656157
## 693 0.18878636
## 694 0.02666667
## 695 0.07509752
## 696 0.00000000
## 697 0.00000000
## 698 0.00000000
## 699 0.00000000
## 700 0.00000000
## 701 0.00000000
## 702 0.00000000
## 703 0.00000000
## 704 0.00000000
## 705 0.03500000
## 706 0.02666667
## 707 0.08644444
## 708 0.17927806
## 709 0.00000000
## 710 0.36272176
## 711 0.46645171
## 712 1.20912696
## 713 0.20458442
## 714 0.70209905
## 715 0.99072785
## 716 0.28816170
## 717 0.27773572
## 718 0.25416747
## 719 0.16241135
## 720 0.07013072
## 721 0.00000000
## 722 0.02000000
## 723 0.05333333
## 724 0.06988304
## 725 0.00000000
## 726 0.00000000
## 727 0.00000000
## 728 0.00000000
## 729 0.00000000
## 730 0.00000000
## 731 0.00000000
## 732 0.00000000
## 733 0.00000000
## 734 0.00000000
## 735 0.00000000
## 736 0.03667519
## 737 0.41402893
## 738 0.15761089
## 739 0.15471774
## 740 0.00000000
## 741 0.06166667
## 742 0.18717949
## 743 0.22528407
## 744 0.15179487
## 745 0.00000000
## 746 0.08625185
## 747 0.02000000
## 748 0.03851852
## 749 0.06666667
## 750 0.00000000
## 751 0.00000000
## 752 0.00000000
## 753 0.00000000
## 754 0.00000000
## 755 0.00000000
## 756 0.00000000
## 757 0.00000000
## 758 0.00000000
## 759 0.00000000
## 760 0.00000000
## 761 0.00000000
## 762 0.00000000
## 763 0.00000000
## 764 0.00000000
## 765 0.00000000
## 766 0.00000000
## 767 0.00000000
## 768 0.00000000
## 769 0.00000000
## 770 0.00000000
## 771 0.00000000
## 772 0.00000000
## 773 0.03000000
## 774 0.00000000
## 775 0.00000000
## 776 0.00000000
## 777 0.00000000
## 778 0.00000000
## 779 0.00000000
## 780 0.00000000
## 781 0.00000000
## 782 0.00000000
## 783 0.00000000
## 784 0.00000000image(matrix(imp, 28, 28))
p_max <- predict(fit_knn, x_test[,col_index])
p_max <- apply(p_max, 1, max)
ind <- which(y_hat_knn != y_test)
ind <- ind[order(p_max[ind], decreasing = TRUE)]
rafalib::mypar(3,4)
for(i in ind[1:12]){
image(matrix(x_test[i,], 28, 28)[, 28:1],
main = paste0("Pr(",y_hat_knn[i],")=",round(p_max[i], 2),
" but is a ",y_test[i]),
xaxt="n", yaxt="n")
}
p_max <- predict(fit_rf, x_test[,col_index])$census
p_max <- p_max / rowSums(p_max)
p_max <- apply(p_max, 1, max)
ind <- which(y_hat_rf != y_test)
ind <- ind[order(p_max[ind], decreasing = TRUE)]
rafalib::mypar(3,4)
for(i in ind[1:12]){
image(matrix(x_test[i,], 28, 28)[, 28:1],
main = paste0("Pr(",y_hat_rf[i],")=",round(p_max[i], 2),
" but is a ",y_test[i]),
xaxt="n", yaxt="n")
}
##Ensemble
p_rf <- predict(fit_rf, x_test[,col_index])$census
p_rf <- p_rf / rowSums(p_rf)
p_knn <- predict(fit_knn, x_test[,col_index])
p <- (p_rf + p_knn)/2
y_pred <- factor(apply(p, 1, which.max)-1)
confusionMatrix(y_pred, y_test)## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2 3 4 5 6 7 8 9
## 0 106 0 0 0 0 1 2 0 0 1
## 1 0 111 0 0 0 1 2 1 1 0
## 2 0 1 91 0 0 0 0 0 1 1
## 3 0 0 0 102 0 4 1 0 2 0
## 4 0 0 0 0 83 0 0 0 1 4
## 5 0 0 0 5 0 83 1 0 0 1
## 6 1 0 1 1 0 2 101 0 3 0
## 7 1 0 4 0 0 0 0 96 0 0
## 8 0 0 0 0 1 2 0 0 89 1
## 9 0 0 0 0 1 1 0 2 0 86
##
## Overall Statistics
##
## Accuracy : 0.948
## 95% CI : (0.9324, 0.9609)
## No Information Rate : 0.112
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9422
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2 Class: 3 Class: 4 Class: 5
## Sensitivity 0.9815 0.9911 0.9479 0.9444 0.9765 0.8830
## Specificity 0.9955 0.9944 0.9967 0.9922 0.9945 0.9923
## Pos Pred Value 0.9636 0.9569 0.9681 0.9358 0.9432 0.9222
## Neg Pred Value 0.9978 0.9989 0.9945 0.9933 0.9978 0.9879
## Prevalence 0.1080 0.1120 0.0960 0.1080 0.0850 0.0940
## Detection Rate 0.1060 0.1110 0.0910 0.1020 0.0830 0.0830
## Detection Prevalence 0.1100 0.1160 0.0940 0.1090 0.0880 0.0900
## Balanced Accuracy 0.9885 0.9927 0.9723 0.9683 0.9855 0.9376
## Class: 6 Class: 7 Class: 8 Class: 9
## Sensitivity 0.9439 0.9697 0.9175 0.9149
## Specificity 0.9910 0.9945 0.9956 0.9956
## Pos Pred Value 0.9266 0.9505 0.9570 0.9556
## Neg Pred Value 0.9933 0.9967 0.9912 0.9912
## Prevalence 0.1070 0.0990 0.0970 0.0940
## Detection Rate 0.1010 0.0960 0.0890 0.0860
## Detection Prevalence 0.1090 0.1010 0.0930 0.0900
## Balanced Accuracy 0.9675 0.9821 0.9565 0.9552ENSEMBLE COM VARIOS MODELOS
models <- c("glm", "lda", "naive_bayes", "svmLinear", "knn", "gamLoess", "multinom", "qda", "rf", "adaboost")
set.seed(1) # use `set.seed(1, sample.kind = "Rounding")` in R 3.6 or later
data("mnist_27")
fits <- lapply(models, function(model){
print(model)
train(y ~ ., method = model, data = mnist_27$train)
}) #Treinar os Modelos## [1] "glm"
## [1] "lda"
## [1] "naive_bayes"
## [1] "svmLinear"
## [1] "knn"
## [1] "gamLoess"## Loading required package: gam## Loading required package: splines## Loading required package: foreach##
## Attaching package: 'foreach'## The following objects are masked from 'package:purrr':
##
## accumulate, when## Loaded gam 1.16.1## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.50205## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.51111## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.50205## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.50205## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53333## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.50205## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.50588## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.50205## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.092316## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored
## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54071## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.092316## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53333## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54071## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54071## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10703## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.094737## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10703## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10723## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.094737## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10723## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54061## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.53555## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53846## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.53333## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.51322## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.092316## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored
## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.43969## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10703## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.094737## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10703## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54071## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.46667## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.40628## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.41379## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.40628## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.4375## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.40628## Warning in gam.lo(data[["lo(x_1, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.089286## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10761## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.094737## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : lowerlimit 0.10761## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : eval 0.57895## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : upperlimit 0.54061## Warning in gam.lo(data[["lo(x_2, span = 0.5, degree = 1)"]], z, w, span =
## 0.5, : extrapolation not allowed with blending## Warning in model.matrix.default(mt, mf, contrasts): non-list contrasts
## argument ignored## [1] "multinom"
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.447413
## final value 340.447361
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 385.009822
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.515756
## final value 340.515703
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 355.491548
## final value 355.334096
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 399.994557
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 355.560397
## final value 355.404046
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 332.977945
## final value 332.776189
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 383.652420
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 333.057166
## final value 332.857086
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 360.815668
## iter 10 value 360.815664
## iter 10 value 360.815664
## final value 360.815664
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 401.991029
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 360.877678
## iter 10 value 360.877674
## iter 10 value 360.877673
## final value 360.877673
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.445113
## final value 353.445099
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 397.620330
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.512471
## final value 353.512457
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 384.705847
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 419.872138
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 384.756753
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 376.311956
## final value 376.311945
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 414.533042
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 376.368612
## final value 376.368601
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 348.115740
## final value 348.115719
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 394.780931
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 348.188604
## final value 348.188583
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 371.294670
## final value 371.294648
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 407.162776
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 371.346769
## final value 371.346747
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.460748
## final value 340.460735
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 385.329076
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.529383
## final value 340.529370
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 369.071134
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 409.102586
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 369.131201
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.789756
## final value 353.789729
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 397.213336
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.855907
## final value 353.855880
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 349.900015
## final value 349.900003
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 395.063948
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 349.969829
## final value 349.969816
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.283330
## final value 340.113998
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 388.970784
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 340.356087
## final value 340.192140
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 356.142461
## iter 10 value 356.142460
## iter 10 value 356.142460
## final value 356.142460
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 400.364737
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 356.210409
## iter 10 value 356.210409
## iter 10 value 356.210408
## final value 356.210408
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 366.460859
## final value 366.457990
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 408.726455
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 366.525723
## final value 366.522943
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 345.681221
## final value 345.681216
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 391.157784
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 345.751777
## final value 345.751772
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 381.951887
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 417.689638
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 382.004339
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 374.392494
## iter 10 value 374.392494
## iter 10 value 374.392494
## final value 374.392494
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 411.255312
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 374.446101
## iter 10 value 374.446101
## iter 10 value 374.446101
## final value 374.446101
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 352.792114
## iter 10 value 352.792111
## iter 10 value 352.792111
## final value 352.792111
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 400.589977
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 352.868524
## iter 10 value 352.868522
## iter 10 value 352.868522
## final value 352.868522
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 367.459858
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 410.336503
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 367.526043
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 361.928018
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 401.510577
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 361.987362
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.664516
## final value 353.664502
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 398.446936
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 353.733429
## final value 353.733415
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 376.662697
## final value 376.662654
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 413.902664
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 376.717948
## final value 376.717905
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 364.624146
## iter 10 value 364.624145
## iter 10 value 364.624145
## final value 364.624145
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 402.783558
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## iter 10 value 364.680086
## iter 10 value 364.680086
## iter 10 value 364.680086
## final value 364.680086
## converged
## # weights: 4 (3 variable)
## initial value 554.517744
## final value 401.160559
## converged
## [1] "qda"
## [1] "rf"
## note: only 1 unique complexity parameters in default grid. Truncating the grid to 1 .
##
## [1] "adaboost"names(fits) <- models
pred <- sapply(fits, function(object)
predict(object, newdata = mnist_27$test)) #Matriz de Previsoes de cada modelo para o conjunto do teste
acc_hat <- sapply(fits, function(fit) min(fit$results$Accuracy))
M=mean(acc_hat) #média das estimativas de precisão do conjunto de treino
ind <- acc_hat >= M #Selecionar só os modelos que contém as estimativas de previsoes acima da média do treino
votes <- rowMeans(pred[,ind] == "7")
y_hat <- ifelse(votes>=0.5, 7, 2)%>%
factor(levels = levels(mnist_27$test$y))
mean(y_hat == mnist_27$test$y)## [1] 0.815confusionMatrix(data=y_hat, reference=mnist_27$test$y) #Matriz de Confusão## Confusion Matrix and Statistics
##
## Reference
## Prediction 2 7
## 2 89 20
## 7 17 74
##
## Accuracy : 0.815
## 95% CI : (0.7541, 0.8663)
## No Information Rate : 0.53
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.628
##
## Mcnemar's Test P-Value : 0.7423
##
## Sensitivity : 0.8396
## Specificity : 0.7872
## Pos Pred Value : 0.8165
## Neg Pred Value : 0.8132
## Prevalence : 0.5300
## Detection Rate : 0.4450
## Detection Prevalence : 0.5450
## Balanced Accuracy : 0.8134
##
## 'Positive' Class : 2
## data("tissue_gene_expression")
dim(tissue_gene_expression$x)## [1] 189 500