Kozmik bir gizemi çözmek için veri bilimi becerilerinize ihtiyaç duyulan 2912 yılına hoş geldiniz. Dört ışıkyılı uzaklıktan bir sinyal aldık ve işler pek iyi görünmüyor.
Titanic uzay gemisi, bir ay önce fırlatılan yıldızlararası bir yolcu gemisidir. Gemide yaklaşık 13.000 yolcu bulunan gemi, göçmenleri güneş sistemimizden yakın yıldızların yörüngesinde bulunan üç yeni yaşanabilir dış gezegene taşımak için ilk yolculuğuna çıktı.
Dikkatsiz uzay gemisi Titanic, Alpha Centauri’nin yörüngesinde ilk varış noktasına (sıcak 55 Cancri E) doğru giderken, bir toz bulutunun içine gizlenmiş bir uzay-zaman anormalliğiyle çarpıştı. Ne yazık ki 1000 yıl önceki adaşı ile benzer bir kaderle karşılaştı. Gemi sağlam kalmasına rağmen yolcuların neredeyse yarısı alternatif bir boyuta taşındı!
library(rmarkdown)
library(explore)
library(tidyverse)
library(readr)
test_c <- read_csv("test_c.csv")
library(readr)
train_c <- read_csv("train_c.csv")
describe_all(train_c)
## # A tibble: 16 × 8
## variable type na na_pct unique min mean max
## <chr> <chr> <int> <dbl> <int> <dbl> <dbl> <dbl>
## 1 PassengerId chr 0 0 8693 NA NA NA
## 2 HomePlanet chr 0 0 3 NA NA NA
## 3 CryoSleep lgl 0 0 2 0 0.35 1
## 4 Destination chr 0 0 4 NA NA NA
## 5 Age dbl 0 0 90 0 28.8 79
## 6 VIP lgl 0 0 2 0 0.02 1
## 7 RoomService dbl 0 0 1273 0 220. 14327
## 8 FoodCourt dbl 0 0 1507 0 448. 29813
## 9 ShoppingMall dbl 0 0 1115 0 170. 23492
## 10 Spa dbl 0 0 1327 0 305. 22408
## 11 VRDeck dbl 0 0 1306 0 298. 24133
## 12 Transported lgl 0 0 2 0 0.5 1
## 13 withgroup dbl 0 0 2 0 0.45 1
## 14 deck chr 0 0 8 NA NA NA
## 15 side chr 0 0 2 NA NA NA
## 16 expense dbl 0 0 2336 0 1441. 35987
describe_all(test_c)
## # A tibble: 15 × 8
## variable type na na_pct unique min mean max
## <chr> <chr> <int> <dbl> <int> <dbl> <dbl> <dbl>
## 1 PassengerId chr 0 0 4277 NA NA NA
## 2 HomePlanet chr 0 0 3 NA NA NA
## 3 CryoSleep lgl 0 0 2 0 0.36 1
## 4 Destination chr 0 0 4 NA NA NA
## 5 Age dbl 0 0 89 0 28.7 79
## 6 VIP lgl 0 0 2 0 0.02 1
## 7 RoomService dbl 0 0 842 0 215. 11567
## 8 FoodCourt dbl 0 0 902 0 429. 25273
## 9 ShoppingMall dbl 0 0 715 0 173. 8292
## 10 Spa dbl 0 0 833 0 296. 19844
## 11 VRDeck dbl 0 0 796 0 305. 22272
## 12 withgroup dbl 0 0 2 0 0.45 1
## 13 deck chr 0 0 8 NA NA NA
## 14 side chr 0 0 2 NA NA NA
## 15 expense dbl 0 0 1437 0 1418. 33666
train_c$HomePlanet <- as.factor(train_c$HomePlanet)
train_c$Destination <- as.factor(train_c$Destination)
train_c$deck <- as.factor(train_c$deck)
train_c$side <- as.factor(train_c$side)
test_c$HomePlanet <- as.factor(test_c$HomePlanet)
test_c$Destination <- as.factor(test_c$Destination)
test_c$deck <- as.factor(test_c$deck)
test_c$side <- as.factor(test_c$side)
library(DataExplorer)
create_report(train_c)
## | | | 0% | |. | 2% | |.. | 5% [global_options] | |... | 7% | |.... | 10% [introduce] | |.... | 12% | |..... | 14% [plot_intro] | |...... | 17% | |....... | 19% [data_structure] | |........ | 21% | |......... | 24% [missing_profile] | |.......... | 26% | |........... | 29% [univariate_distribution_header] | |........... | 31% | |............ | 33% [plot_histogram] | |............. | 36% | |.............. | 38% [plot_density] | |............... | 40% | |................ | 43% [plot_frequency_bar] | |................. | 45% | |.................. | 48% [plot_response_bar] | |.................. | 50% | |................... | 52% [plot_with_bar] | |.................... | 55% | |..................... | 57% [plot_normal_qq] | |...................... | 60% | |....................... | 62% [plot_response_qq] | |........................ | 64% | |......................... | 67% [plot_by_qq] | |.......................... | 69% | |.......................... | 71% [correlation_analysis] | |........................... | 74% | |............................ | 76% [principal_component_analysis] | |............................. | 79% | |.............................. | 81% [bivariate_distribution_header] | |............................... | 83% | |................................ | 86% [plot_response_boxplot] | |................................. | 88% | |................................. | 90% [plot_by_boxplot] | |.................................. | 93% | |................................... | 95% [plot_response_scatterplot] | |.................................... | 98% | |.....................................| 100% [plot_by_scatterplot]
## "C:/Program Files/RStudio/resources/app/bin/quarto/bin/tools/pandoc" +RTS -K512m -RTS "C:\EKONOM~1\REPORT~1.MD" --to html4 --from markdown+autolink_bare_uris+tex_math_single_backslash --output pandoc628c2dee6805.html --lua-filter "C:\Users\ahmat\AppData\Local\R\win-library\4.3\rmarkdown\rmarkdown\lua\pagebreak.lua" --lua-filter "C:\Users\ahmat\AppData\Local\R\win-library\4.3\rmarkdown\rmarkdown\lua\latex-div.lua" --embed-resources --standalone --variable bs3=TRUE --section-divs --table-of-contents --toc-depth 6 --template "C:\Users\ahmat\AppData\Local\R\win-library\4.3\rmarkdown\rmd\h\default.html" --no-highlight --variable highlightjs=1 --variable theme=yeti --mathjax --variable "mathjax-url=https://mathjax.rstudio.com/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML" --include-in-header "C:\Users\ahmat\AppData\Local\Temp\RtmpAhOgvK\rmarkdown-str628c641a7e4e.html"
model = lm(Transported ~ ., data = train_c[, 2:16])
summary(model)
##
## Call:
## lm(formula = Transported ~ ., data = train_c[, 2:16])
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.42802 -0.30952 -0.03188 0.28870 1.79172
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.147e-01 4.000e-02 7.866 4.09e-15 ***
## HomePlanetEuropa 2.150e-01 2.883e-02 7.457 9.72e-14 ***
## HomePlanetMars 9.617e-02 1.475e-02 6.522 7.33e-11 ***
## CryoSleepTRUE 3.811e-01 1.152e-02 33.082 < 2e-16 ***
## DestinationPSO J318.5-22 -4.449e-02 1.803e-02 -2.468 0.013598 *
## DestinationTRAPPIST-1e -5.140e-02 1.133e-02 -4.538 5.76e-06 ***
## DestinationTRAPPIST-le 7.513e-03 3.142e-02 0.239 0.811033
## Age -2.182e-03 3.201e-04 -6.815 1.00e-11 ***
## VIPTRUE -3.757e-02 2.974e-02 -1.263 0.206555
## RoomService -1.176e-04 7.055e-06 -16.674 < 2e-16 ***
## FoodCourt 4.281e-05 3.054e-06 14.017 < 2e-16 ***
## ShoppingMall 7.969e-05 7.457e-06 10.686 < 2e-16 ***
## Spa -8.675e-05 4.109e-06 -21.110 < 2e-16 ***
## VRDeck -8.280e-05 4.115e-06 -20.120 < 2e-16 ***
## withgroup 1.958e-02 9.293e-03 2.107 0.035134 *
## deckB 1.107e-01 2.874e-02 3.852 0.000118 ***
## deckC 1.477e-01 2.900e-02 5.094 3.58e-07 ***
## deckD 5.090e-02 3.478e-02 1.463 0.143372
## deckE 8.494e-03 3.622e-02 0.235 0.814567
## deckF 1.031e-01 3.709e-02 2.778 0.005475 **
## deckG 5.791e-02 3.866e-02 1.498 0.134218
## deckT 6.610e-02 1.815e-01 0.364 0.715784
## sideS 8.542e-02 8.629e-03 9.899 < 2e-16 ***
## expense NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.4012 on 8670 degrees of freedom
## Multiple R-squared: 0.3578, Adjusted R-squared: 0.3561
## F-statistic: 219.5 on 22 and 8670 DF, p-value: < 2.2e-16
library(caTools)
set.seed(123)
split = sample.split(train_c$Transported, SplitRatio = 0.75)
train_train = subset(train_c, split == TRUE)
train_test = subset(train_c, split == FALSE)
regresyon = lm(Transported ~ ., data = train_train[, -c(1)])
summary(regresyon)
##
## Call:
## lm(formula = Transported ~ ., data = train_train[, -c(1)])
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.50155 -0.31213 -0.03018 0.29171 1.76723
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.482e-01 4.626e-02 7.527 5.89e-14 ***
## HomePlanetEuropa 1.971e-01 3.358e-02 5.871 4.55e-09 ***
## HomePlanetMars 8.600e-02 1.699e-02 5.061 4.29e-07 ***
## CryoSleepTRUE 3.772e-01 1.328e-02 28.402 < 2e-16 ***
## DestinationPSO J318.5-22 -6.107e-02 2.077e-02 -2.940 0.00330 **
## DestinationTRAPPIST-1e -5.426e-02 1.309e-02 -4.144 3.46e-05 ***
## DestinationTRAPPIST-le -9.053e-03 3.649e-02 -0.248 0.80407
## Age -2.339e-03 3.725e-04 -6.279 3.63e-10 ***
## VIPTRUE -1.616e-02 3.384e-02 -0.478 0.63290
## RoomService -1.175e-04 7.790e-06 -15.088 < 2e-16 ***
## FoodCourt 3.905e-05 3.467e-06 11.262 < 2e-16 ***
## ShoppingMall 8.282e-05 8.415e-06 9.842 < 2e-16 ***
## Spa -8.543e-05 4.704e-06 -18.160 < 2e-16 ***
## VRDeck -8.216e-05 4.777e-06 -17.197 < 2e-16 ***
## withgroup 1.601e-02 1.076e-02 1.488 0.13688
## deckB 1.049e-01 3.290e-02 3.188 0.00144 **
## deckC 1.441e-01 3.318e-02 4.343 1.43e-05 ***
## deckD 3.223e-02 3.974e-02 0.811 0.41744
## deckE -2.463e-02 4.213e-02 -0.585 0.55884
## deckF 8.346e-02 4.290e-02 1.945 0.05176 .
## deckG 4.007e-02 4.472e-02 0.896 0.37029
## deckT 5.680e-02 1.822e-01 0.312 0.75526
## sideS 8.895e-02 9.984e-03 8.909 < 2e-16 ***
## expense NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.4013 on 6497 degrees of freedom
## Multiple R-squared: 0.358, Adjusted R-squared: 0.3558
## F-statistic: 164.7 on 22 and 6497 DF, p-value: < 2.2e-16
reg_tahmin = predict(regresyon, newdata = train_test[, -c(1,12)])
reg_transported_tahmin = ifelse(reg_tahmin > 0.5, 1, 0)
transported_gercek = ifelse(train_test[12] == TRUE, 1, 0)
cm = table(transported_gercek, reg_transported_tahmin)
cm
## reg_transported_tahmin
## transported_gercek 0 1
## 0 902 177
## 1 323 771
(902 + 771)/(902 + 771 + 323 + 177)
## [1] 0.7699034
reg_tahmin = predict(regresyon, newdata = train_train[, -c(1,12)])
reg_transported_tahmin = ifelse(reg_tahmin > 0.5, 1, 0)
transported_gercek = ifelse(train_train[12] == TRUE, 1, 0)
cm = table(transported_gercek, reg_transported_tahmin)
cm
## reg_transported_tahmin
## transported_gercek 0 1
## 0 2682 554
## 1 946 2338
(2682 + 2338)/(2682 + 2338 + 554 + 946)
## [1] 0.7699387
reg_tahmin_bd = predict(model, newdata = test_c[, -c(1)])
reg_transported_test_c_tahmin = ifelse(reg_tahmin_bd > 0.5, TRUE, FALSE)
transported = as.character(reg_transported_test_c_tahmin)
PassengerId = test_c$PassengerId
Transported=as.vector(transported)
submission_regresyon = cbind(PassengerId, Transported)
submission_regresyon = as.data.frame(submission_regresyon)
library(stringr)
submission_regresyon$transported <- str_to_title(submission_regresyon$Transported)
write.csv(submission_regresyon, "siniftahmini.csv", row.names = FALSE, quote = FALSE)
train_c_log = train_c %>% mutate_at(c(5, 7:11, 16), ~ log(1 + .))
test_c_log = test_c %>% mutate_at(c(5, 7:11, 15), ~ log(1 + .))
modellog = lm(Transported ~ 1 + HomePlanet + Destination + deck + side + CryoSleep + Age + RoomService + FoodCourt + ShoppingMall + Spa + VRDeck, data = train_c_log)
reg_tahmin_log = predict(modellog, newdata = test_c_log[, -c(1)])
reg_transported_test_c_tahmin_log = ifelse(reg_tahmin_log > 0.5, TRUE, FALSE)
head(reg_transported_test_c_tahmin_log)
## 1 2 3 4 5 6
## TRUE FALSE TRUE TRUE TRUE FALSE
Transported = as.character(reg_transported_test_c_tahmin_log)
PassengerId = test_c_log$PassengerId
Transported = as.vector(Transported)
submission_regresyon_log = cbind(PassengerId, Transported)
submission_regresyon_log =as.data.frame(submission_regresyon_log)
submission_regresyon_log$Transported = str_to_title(submission_regresyon_log$Transported)
write.csv(submission_regresyon_log,"submissyon_regresyon_log.csv", row.names = FALSE,quote = FALSE)
model = lm(y ~ x1 + x2 + x3, data = df)
logistic = glm(formula = Transported ~ .,
family = binomial,
data = train_train[, -c(1)])
summary(logistic)
##
## Call:
## glm(formula = Transported ~ ., family = binomial, data = train_train[,
## -c(1)])
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.607e-01 3.588e-01 -1.284 0.19917
## HomePlanetEuropa 1.556e+00 2.674e-01 5.819 5.91e-09 ***
## HomePlanetMars 5.323e-01 1.103e-01 4.827 1.39e-06 ***
## CryoSleepTRUE 1.285e+00 9.185e-02 13.987 < 2e-16 ***
## DestinationPSO J318.5-22 -5.125e-01 1.297e-01 -3.950 7.80e-05 ***
## DestinationTRAPPIST-1e -4.904e-01 9.441e-02 -5.194 2.05e-07 ***
## DestinationTRAPPIST-le -1.471e-01 2.436e-01 -0.604 0.54580
## Age -8.481e-03 2.461e-03 -3.446 0.00057 ***
## VIPTRUE -1.650e-01 2.916e-01 -0.566 0.57149
## RoomService -1.730e-03 1.136e-04 -15.225 < 2e-16 ***
## FoodCourt 4.581e-04 4.460e-05 10.271 < 2e-16 ***
## ShoppingMall 5.289e-04 7.559e-05 6.997 2.61e-12 ***
## Spa -1.969e-03 1.164e-04 -16.913 < 2e-16 ***
## VRDeck -1.902e-03 1.161e-04 -16.392 < 2e-16 ***
## withgroup 1.206e-01 7.192e-02 1.677 0.09346 .
## deckB 1.153e+00 2.912e-01 3.961 7.47e-05 ***
## deckC 2.481e+00 3.325e-01 7.463 8.48e-14 ***
## deckD 6.978e-01 3.262e-01 2.139 0.03242 *
## deckE 1.029e-01 3.378e-01 0.305 0.76065
## deckF 7.739e-01 3.420e-01 2.263 0.02363 *
## deckG 3.766e-01 3.511e-01 1.073 0.28345
## deckT -1.184e-01 1.866e+00 -0.063 0.94942
## sideS 5.988e-01 6.663e-02 8.987 < 2e-16 ***
## expense NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 9038.3 on 6519 degrees of freedom
## Residual deviance: 5600.2 on 6497 degrees of freedom
## AIC: 5646.2
##
## Number of Fisher Scoring iterations: 7
logistic_tahmin = predict(logistic, newdata = train_test[, -c(1,12)])
head(logistic_tahmin)
## 1 2 3 4 5 6
## -1.1175750 -11.0810548 0.9556805 -1.6375160 -0.4294310 -1.6887335
logistic_transported_tahmin = ifelse(logistic_tahmin > 0.5, 1, 0)
head(logistic_transported_tahmin)
## 1 2 3 4 5 6
## 0 0 1 0 0 0
transported_gercek = ifelse(train_test[12] == TRUE, 1, 0)
library(tidymodels)
result = data.frame(cbind(transported_gercek, logistic_transported_tahmin))
result$Transported = as.factor(result$Transported)
result$logistic_transported_tahmin = as.factor(result$logistic_transported_tahmin)
accuracy(result, truth = Transported, estimate = logistic_transported_tahmin)
## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 accuracy binary 0.786
conf_mat(result, truth = Transported, estimate = logistic_transported_tahmin)
## Truth
## Prediction 0 1
## 0 918 305
## 1 161 789
library(caret)
cm = table(transported_gercek, logistic_transported_tahmin)
cm
## logistic_transported_tahmin
## transported_gercek 0 1
## 0 918 161
## 1 305 789
(918 +789)/(918 +161 + 305 +789)
## [1] 0.7855499
confusionMatrix(as.factor(transported_gercek), as.factor(logistic_transported_tahmin))
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 918 161
## 1 305 789
##
## Accuracy : 0.7855
## 95% CI : (0.7677, 0.8026)
## No Information Rate : 0.5628
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.5715
##
## Mcnemar's Test P-Value : 3.488e-11
##
## Sensitivity : 0.7506
## Specificity : 0.8305
## Pos Pred Value : 0.8508
## Neg Pred Value : 0.7212
## Prevalence : 0.5628
## Detection Rate : 0.4225
## Detection Prevalence : 0.4965
## Balanced Accuracy : 0.7906
##
## 'Positive' Class : 0
##
(918+789)/(918+789+161+305)
## [1] 0.7855499
logistic_bd = glm(formula = Transported ~ .,
family = binomial,
data = train_c[, -c(1)])
logistic_tahmin_bd = predict(logistic_bd, newdata = test_c[, -c(1)])
logistic_transported_test_c_tahmin = ifelse(logistic_tahmin_bd >0.5, TRUE, FALSE )
Transported = as.character(logistic_transported_test_c_tahmin)
PassengerId = test_c$PassengerId
Transported=as.vector(Transported)
submission_logistic = cbind(PassengerId, Transported)
submission_logistic = as.data.frame(submission_logistic)
submission_logistic$Transported = str_to_title(submission_logistic$Transported)
write.csv(submission_logistic, "submission_logistic.csv", row.names =FALSE, quote=FALSE)
logistic_bd = glm(formula = Transported ~ .,
family = binomial,
data = train_c_log[, -c(1)])
logistic_tahmin_bd = predict(logistic_bd, newdata = test_c_log[, -c(1)])
logistic_transported_test_c_tahmin = ifelse(logistic_tahmin_bd >0.5, TRUE, FALSE )
Transported = as.character(logistic_transported_test_c_tahmin)
PassengerId = test_c$PassengerId
Transported=as.vector(Transported)
submission_llogistic = cbind(PassengerId, Transported)
submission_llogistic = as.data.frame(submission_logistic)
submission_llogistic$Transported = str_to_title(submission_logistic$Transported)
write.csv(submission_logistic, "submission_llogistic.csv", row.names =FALSE, quote=FALSE)
Naive Bayes, olasılık temelli bir makine öğrenmesi algoritmasıdır. Naive Bayes sınıflandırıcısı, basit ama güçlü bir sınıflandırma yöntemidir.
Naive Bayes algoritmasının temel özellikleri ve çalışma prensibi şu şekilde açıklanabilir:
Naive Bayes, basitliği, hızlı çalışması ve iyi performansı nedeniyle makine öğrenmesi uygulamalarında sıklıkla tercih edilen bir algoritmadır.
library(e1071)
fit_nb = naiveBayes(Transported ~ ., data =train_train[, -1])
fit_nb
##
## Naive Bayes Classifier for Discrete Predictors
##
## Call:
## naiveBayes.default(x = X, y = Y, laplace = laplace)
##
## A-priori probabilities:
## Y
## FALSE TRUE
## 0.496319 0.503681
##
## Conditional probabilities:
## HomePlanet
## Y Earth Europa Mars
## FALSE 0.6236094 0.1739802 0.2024104
## TRUE 0.4631547 0.3279537 0.2088916
##
## CryoSleep
## Y FALSE TRUE
## FALSE 0.8702101 0.1297899
## TRUE 0.4336175 0.5663825
##
## Destination
## Y 55 Cancri e PSO J318.5-22 TRAPPIST-1e TRAPPIST-le
## FALSE 0.16069221 0.09363412 0.72404203 0.02163164
## TRUE 0.24969549 0.09043849 0.64007308 0.01979294
##
## Age
## Y [,1] [,2]
## FALSE 30.01092 13.45222
## TRUE 27.85527 14.87646
##
## VIP
## Y FALSE TRUE
## FALSE 0.97126082 0.02873918
## TRUE 0.98142509 0.01857491
##
## RoomService
## Y [,1] [,2]
## FALSE 402.20365 916.6547
## TRUE 56.93484 246.8105
##
## FoodCourt
## Y [,1] [,2]
## FALSE 396.2923 1258.123
## TRUE 514.8018 1918.620
##
## ShoppingMall
## Y [,1] [,2]
## FALSE 160.2250 432.1103
## TRUE 182.7135 748.5867
##
## Spa
## Y [,1] [,2]
## FALSE 561.16193 1554.2636
## TRUE 61.11571 264.0556
##
## VRDeck
## Y [,1] [,2]
## FALSE 533.98733 1542.8703
## TRUE 69.07186 295.0542
##
## withgroup
## Y [,1] [,2]
## FALSE 0.3952410 0.4889780
## TRUE 0.4960414 0.5000605
##
## deck
## Y A B C D E
## FALSE 0.0296662546 0.0540791100 0.0574783684 0.0661310260 0.1344252163
## TRUE 0.0310596833 0.1385505481 0.1199756395 0.0490255786 0.0676004872
## deck
## Y F G T
## FALSE 0.3637206428 0.2932632880 0.0012360939
## TRUE 0.2828867235 0.3105968331 0.0003045067
##
## side
## Y P S
## FALSE 0.5571693 0.4428307
## TRUE 0.4448843 0.5551157
##
## expense
## Y [,1] [,2]
## FALSE 2053.8702 3224.219
## TRUE 884.6376 2307.962
pred_nb = predict(fit_nb, newdata = train_test[, -c(1, 12)], type = "raw") %>%
data.frame()
head(pred_nb)
## FALSE. TRUE.
## 1 0.22238032 7.776197e-01
## 2 1.00000000 8.663595e-135
## 3 0.02729931 9.727007e-01
## 4 0.59276719 4.072328e-01
## 5 0.02410104 9.758990e-01
## 6 0.38468138 6.153186e-01
Transported_pred_nb = ifelse(pred_nb$TRUE. > 0.5, 1, 0)
head(Transported_pred_nb)
## [1] 1 0 1 0 1 1
Transported_train_test = ifelse(train_test[12] == TRUE, 1, 0)
head(Transported_train_test)
## Transported
## [1,] 1
## [2,] 0
## [3,] 1
## [4,] 0
## [5,] 0
## [6,] 0
cm = table(Transported_train_test, Transported_pred_nb)
cm
## Transported_pred_nb
## Transported_train_test 0 1
## 0 480 599
## 1 76 1018
(480 + 1018)/(480 +599 +76 +1018)
## [1] 0.6893695
nb = naiveBayes(Transported ~ ., data = train_c[, -1])
pred_nb = predict(nb, newdata = test_c, type = "raw") %>% data.frame()
Transported_pred_nb = ifelse(pred_nb$TRUE. > 0.5, TRUE, FALSE)
Transported = as.character(Transported_pred_nb)
PassengerId = test_c$PassengerId
Transported = as.vector(Transported)
sample_submission = cbind(PassengerId, Transported)
sample_submission = as.data.frame(sample_submission)
sample_submission$Transported = str_to_title(sample_submission$Transported)
write.csv(sample_submission, "sub_nb.csv", row.names = FALSE, quote = FALSE)
SVM (Support Vector Machine) Radyal, makine öğrenmesinde kullanılan önemli bir algoritma türüdür. Radyal tabanlı SVM, çekirdek fonksiyonu olarak radyal tabanlı fonksiyonu (Radial Basis Function - RBF) kullanan bir SVM modelidir.
Radyal tabanlı SVM’nin temel özellikleri ve işleyişi şu şekilde açıklanabilir:
Radyal tabanlı SVM, görüntü sınıflandırması, doğal dil işleme, biyoinformatik ve finans gibi çeşitli alanlarda yaygın olarak kullanılmaktadır. Luck only smiles once SVM (Support Vector Machine) Radyal, makine öğrenmesinde kullanılan önemli bir algoritma türüdür. Radyal tabanlı SVM, çekirdek fonksiyonu olarak radyal tabanlı fonksiyonu (Radial Basis Function - RBF) kullanan bir SVM modelidir.
Radyal tabanlı SVM’nin temel özellikleri ve işleyişi şu şekilde açıklanabilir:
Radyal tabanlı SVM, görüntü sınıflandırması, doğal dil işleme, biyoinformatik ve finans gibi çeşitli alanlarda yaygın olarak kullanılmaktadır.
fit_svm = svm(Transported ~ ., data = train_train[, -1],
type= 'C-classification',
kernel = 'linear')
preds = predict(fit_svm, newdata = train_test[, -c(1, 12)], type = "raw") %>%
data.frame()
head(preds)
## .
## 1 FALSE
## 2 FALSE
## 3 TRUE
## 4 FALSE
## 5 FALSE
## 6 FALSE
Transported_pred_svm = ifelse(preds$. == TRUE, 1, 0)
cm = table(Transported_train_test, Transported_pred_svm)
cm
## Transported_pred_svm
## Transported_train_test 0 1
## 0 836 243
## 1 180 914
(836 + 914)/(836 +243 +180 + 914)
## [1] 0.8053382
fit_svm = svm(Transported ~ ., data = train_c[, -1],
type= 'C-classification',
kernel = 'linear')
preds = predict(fit_svm, newdata = test_c, type ="raw")%>%
data.frame()
head(preds)
## .
## 1 TRUE
## 2 FALSE
## 3 TRUE
## 4 TRUE
## 5 TRUE
## 6 TRUE
Transported_pred_svm = ifelse(preds$. == TRUE, TRUE, FALSE)
Transported = as.character(Transported_pred_svm)
PassengerId = test_c$PassengerId
Transported = as.vector(Transported)
sample_submission = cbind(PassengerId, Transported)
sample_submission = as.data.frame(sample_submission)
sample_submission$Transported = str_to_title(sample_submission$Transported)
write.csv(sample_submission, "sub_svm.csv", row.names = FALSE, quote = FALSE)
fit_svm = svm(Transported ~ ., data = train_train[, -1],
type= 'C-classification',
kernel = 'radial')
preds = predict(fit_svm, newdata = train_test[, -c(1, 12)], type ="raw")%>%
data.frame()
Transported_pred_svm = ifelse(preds$. == TRUE, 1, 0)
cm = table(Transported_train_test, Transported_pred_svm)
cm
## Transported_pred_svm
## Transported_train_test 0 1
## 0 849 230
## 1 202 892
(849 +892)/(892 +849 +202 +230)
## [1] 0.8011965
fit_svm = svm(Transported ~ ., data = train_c[, -1],
type= 'C-classification',
kernel = 'radial')
preds = predict(fit_svm, newdata = test_c, type ="raw")%>%
data.frame()
Transported_pred_svm = ifelse(preds$. == TRUE, TRUE, FALSE)
Transported = as.character(Transported_pred_svm)
PassengerId = test_c$PassengerId
Transported = as.vector(Transported)
sample_submission = cbind(PassengerId, Transported)
sample_submission = as.data.frame(sample_submission)
sample_submission$Transported = str_to_title(sample_submission$Transported)
write.csv(sample_submission, "sub_svm_radial.csv", row.names = FALSE, quote = FALSE)
P = ggplot(train_train, aes(x=HomePlanet, y=deck, color=factor(Transported))) +
geom_point(aes(shape=factor(Transported)), size=3) +
scale_color_viridis_d() + labs(title = "", x="HomePlanet", y="deck") + theme_minimal() + theme(legend.position ="top")
P
library(rpart)
library(rpart.plot)
library(randomForest)
library(caret)
fit_tree = rpart::rpart(Transported ~ ., data = train_train[, -1])
summary(fit_tree)
## Call:
## rpart::rpart(formula = Transported ~ ., data = train_train[,
## -1])
## n= 6520
##
## CP nsplit rel error xerror xstd
## 1 0.23321359 0 1.0000000 1.0001811 0.0001977604
## 2 0.04905507 1 0.7667864 0.7671015 0.0103212807
## 3 0.02807107 2 0.7177313 0.7180892 0.0090936774
## 4 0.02784643 3 0.6896603 0.7039381 0.0095638265
## 5 0.01770479 4 0.6618138 0.6661699 0.0097007659
## 6 0.01381587 5 0.6441090 0.6496207 0.0100644975
## 7 0.01194576 6 0.6302932 0.6389139 0.0102048399
## 8 0.01164524 7 0.6183474 0.6310762 0.0102269497
## 9 0.01000000 8 0.6067022 0.6242632 0.0102752753
##
## Variable importance
## expense CryoSleep FoodCourt Spa VRDeck RoomService
## 23 17 13 12 11 9
## deck HomePlanet ShoppingMall Age Destination
## 5 5 2 1 1
##
## Node number 1: 6520 observations, complexity param=0.2332136
## mean=0.503681, MSE=0.2499865
## left son=2 (3795 obs) right son=3 (2725 obs)
## Primary splits:
## expense < 0.5 to the right, improve=0.2332136, (0 missing)
## CryoSleep < 0.5 to the left, improve=0.2095384, (0 missing)
## RoomService < 0.5 to the right, improve=0.1254121, (0 missing)
## Spa < 0.5 to the right, improve=0.1142180, (0 missing)
## VRDeck < 0.5 to the right, improve=0.1084074, (0 missing)
## Surrogate splits:
## CryoSleep < 0.5 to the left, agree=0.932, adj=0.837, (0 split)
## Spa < 0.5 to the right, agree=0.784, adj=0.484, (0 split)
## FoodCourt < 0.5 to the right, agree=0.770, adj=0.451, (0 split)
## VRDeck < 0.5 to the right, agree=0.768, adj=0.444, (0 split)
## RoomService < 0.5 to the right, agree=0.757, adj=0.419, (0 split)
##
## Node number 2: 3795 observations, complexity param=0.02807107
## mean=0.2990777, MSE=0.2096302
## left son=4 (3243 obs) right son=5 (552 obs)
## Primary splits:
## FoodCourt < 1331 to the left, improve=0.05751186, (0 missing)
## ShoppingMall < 627.5 to the left, improve=0.04530804, (0 missing)
## RoomService < 365.5 to the right, improve=0.04440387, (0 missing)
## Spa < 257.5 to the right, improve=0.03347453, (0 missing)
## VRDeck < 721 to the right, improve=0.02290457, (0 missing)
## Surrogate splits:
## expense < 5981 to the left, agree=0.885, adj=0.210, (0 split)
## deck splits as RRRLLLLL, agree=0.884, adj=0.203, (0 split)
## HomePlanet splits as LRL, agree=0.878, adj=0.159, (0 split)
## Spa < 8955.5 to the left, agree=0.856, adj=0.009, (0 split)
## VRDeck < 11692 to the left, agree=0.856, adj=0.009, (0 split)
##
## Node number 3: 2725 observations, complexity param=0.04905507
## mean=0.7886239, MSE=0.1666963
## left son=6 (1448 obs) right son=7 (1277 obs)
## Primary splits:
## deck splits as RRRRLRL-, improve=0.17601740, (0 missing)
## HomePlanet splits as LRR, improve=0.12704810, (0 missing)
## Destination splits as RLLR, improve=0.02774089, (0 missing)
## CryoSleep < 0.5 to the left, improve=0.02268236, (0 missing)
## side splits as LR, improve=0.01498707, (0 missing)
## Surrogate splits:
## HomePlanet splits as LRR, agree=0.934, adj=0.860, (0 split)
## Age < 24.5 to the left, agree=0.626, adj=0.201, (0 split)
## Destination splits as RLLR, agree=0.596, adj=0.137, (0 split)
## withgroup < 0.5 to the left, agree=0.583, adj=0.110, (0 split)
## VIP < 0.5 to the left, agree=0.538, adj=0.014, (0 split)
##
## Node number 4: 3243 observations, complexity param=0.02784643
## mean=0.2537774, MSE=0.1893744
## left son=8 (2577 obs) right son=9 (666 obs)
## Primary splits:
## ShoppingMall < 541.5 to the left, improve=0.07390355, (0 missing)
## RoomService < 365.5 to the right, improve=0.03464407, (0 missing)
## Spa < 240.5 to the right, improve=0.03327259, (0 missing)
## VRDeck < 114 to the right, improve=0.02784287, (0 missing)
## expense < 2867.5 to the right, improve=0.01811461, (0 missing)
## Surrogate splits:
## expense < 18644 to the left, agree=0.795, adj=0.003, (0 split)
##
## Node number 5: 552 observations, complexity param=0.01770479
## mean=0.5652174, MSE=0.2457467
## left son=10 (123 obs) right son=11 (429 obs)
## Primary splits:
## Spa < 1372.5 to the right, improve=0.21272970, (0 missing)
## VRDeck < 1063.5 to the right, improve=0.17089500, (0 missing)
## expense < 5395 to the right, improve=0.06611166, (0 missing)
## FoodCourt < 2513 to the left, improve=0.02780045, (0 missing)
## deck splits as LLRLLRRL, improve=0.02705865, (0 missing)
## Surrogate splits:
## expense < 12647 to the right, agree=0.790, adj=0.057, (0 split)
## Age < 13.5 to the left, agree=0.779, adj=0.008, (0 split)
## RoomService < 3895.5 to the right, agree=0.779, adj=0.008, (0 split)
##
## Node number 6: 1448 observations
## mean=0.6277624, MSE=0.2336768
##
## Node number 7: 1277 observations
## mean=0.9710258, MSE=0.02813466
##
## Node number 8: 2577 observations, complexity param=0.01194576
## mean=0.193636, MSE=0.1561411
## left son=16 (2067 obs) right son=17 (510 obs)
## Primary splits:
## FoodCourt < 456.5 to the left, improve=0.04838893, (0 missing)
## expense < 1447.5 to the right, improve=0.04016842, (0 missing)
## HomePlanet splits as RLL, improve=0.02539308, (0 missing)
## Spa < 537.5 to the right, improve=0.01895890, (0 missing)
## RoomService < 400.5 to the right, improve=0.01706521, (0 missing)
## Surrogate splits:
## expense < 12373 to the left, agree=0.804, adj=0.008, (0 split)
## Spa < 13650 to the left, agree=0.803, adj=0.006, (0 split)
## VRDeck < 10123.5 to the left, agree=0.802, adj=0.002, (0 split)
## deck splits as LLLLLLLR, agree=0.802, adj=0.002, (0 split)
##
## Node number 9: 666 observations
## mean=0.4864865, MSE=0.2498174
##
## Node number 10: 123 observations
## mean=0.1382114, MSE=0.119109
##
## Node number 11: 429 observations, complexity param=0.01381587
## mean=0.6876457, MSE=0.2147891
## left son=22 (143 obs) right son=23 (286 obs)
## Primary splits:
## VRDeck < 611 to the right, improve=0.24438400, (0 missing)
## Spa < 225 to the right, improve=0.05300377, (0 missing)
## FoodCourt < 3119.5 to the left, improve=0.05168044, (0 missing)
## side splits as LR, improve=0.04004566, (0 missing)
## RoomService < 1719.5 to the right, improve=0.03930897, (0 missing)
## Surrogate splits:
## expense < 6032 to the right, agree=0.702, adj=0.105, (0 split)
## Age < 53.5 to the right, agree=0.674, adj=0.021, (0 split)
## FoodCourt < 12128.5 to the right, agree=0.671, adj=0.014, (0 split)
##
## Node number 16: 2067 observations
## mean=0.1504596, MSE=0.1278215
##
## Node number 17: 510 observations, complexity param=0.01164524
## mean=0.3686275, MSE=0.2327413
## left son=34 (204 obs) right son=35 (306 obs)
## Primary splits:
## expense < 1447.5 to the right, improve=0.15990760, (0 missing)
## VRDeck < 86.5 to the right, improve=0.10060710, (0 missing)
## HomePlanet splits as RLL, improve=0.07952538, (0 missing)
## Spa < 500 to the right, improve=0.07353369, (0 missing)
## deck splits as LLLLRRRL, improve=0.05075444, (0 missing)
## Surrogate splits:
## HomePlanet splits as RLL, agree=0.873, adj=0.681, (0 split)
## deck splits as LLLLRRRL, agree=0.839, adj=0.598, (0 split)
## VRDeck < 213.5 to the right, agree=0.818, adj=0.544, (0 split)
## Spa < 219.5 to the right, agree=0.792, adj=0.480, (0 split)
## FoodCourt < 907 to the right, agree=0.722, adj=0.304, (0 split)
##
## Node number 22: 143 observations
## mean=0.3636364, MSE=0.231405
##
## Node number 23: 286 observations
## mean=0.8496503, MSE=0.1277446
##
## Node number 34: 204 observations
## mean=0.1323529, MSE=0.1148356
##
## Node number 35: 306 observations
## mean=0.5261438, MSE=0.2493165
rpart.plot(fit_tree)
preds= predict(fit_tree, newdata = train_test[, -c(1,12)]) %>%
data.frame()
head(preds)
## .
## 1 0.1504596
## 2 0.1382114
## 3 0.8496503
## 4 0.1504596
## 5 0.4864865
## 6 0.1504596
Transported_pred_tree = ifelse(preds$. >0.5, 1, 0)
cm = table(Transported_train_test, Transported_pred_tree)
cm
## Transported_pred_tree
## Transported_train_test 0 1
## 0 807 272
## 1 237 857
(807 +857)/(807 +857 +272 +237)
## [1] 0.7657616
fit_tree = rpart::rpart(Transported ~ ., data = train_c[, -1])
preds= predict(fit_tree, newdata = test_c) %>%
data.frame()
Transported_pred_tree = ifelse(preds$. > 0.5, TRUE, FALSE)
Transported = as.character(Transported_pred_tree)
PassengerId = test_c$PassengerId
Transported = as.vector(Transported)
sample_submission = cbind(PassengerId, Transported)
sample_submission = as.data.frame(sample_submission)
sample_submission$Transported = str_to_title(sample_submission$Transported)
write.csv(sample_submission, "sub_tree.csv", row.names = FALSE, quote = FALSE)
İşte Random Forest algoritmasının Türkçe açıklaması:
Random Forest, sınıflandırma ve regresyon görevleri için kullanılan popüler bir makine öğrenmesi algoritmasıdır. Birden çok karar ağacının birleştirilmesi yoluyla modelin genel tahmin performansını ve güvenilirliğini artıran bir topluluk öğrenme yöntemidir.
Daha ayrıntılı olarak:
Karar Ağaçları: Random Forest, bireysel karar ağaçları üzerine kurulur. Her karar ağacı, girdi verilerinin özellik değerlerine dayalı bir dizi bölme kararı izleyerek tahminler yapan hiyerarşik bir modeldir.
Topluluk Öğrenmesi: Random Forest, birden çok karar ağacını bir araya getirerek bir topluluk modeli oluşturur. Çok sayıda karar ağacının tahminlerini birleştirerek, tek bir karar ağacından daha iyi performans elde edebilir.
Rastgelelik:
Oylama/Ortalama: Sınıflandırma görevleri için, Random Forest bireysel karar ağaçlarının tahminleri arasından çoğunluk oyuyla tahmin yapar. Regresyon görevleri için ise tahminlerin ortalamasını alır.
Dayanıklılık: Bireysel karar ağaçlarının eğitim sürecinde eklenen rastgelelik, genel Random Forest modelini aşırı uyuma, veri gürültüsüne ve aykırı değerlere karşı daha dayanıklı hale getirir.
Random Forest’ın başlıca avantajları: - Hem sayısal hem de kategorik değişkenleri işleyebilir - Yüksek boyutlu, büyük veri kümeleri üzerinde iyi performans gösterir - Özellik önem derecelerini sağlar - Nispeten kolay ayarlanabilir ve yorumlanabilir
Random Forest, esnekliği ve güçlü performansı nedeniyle görüntü tanıma, doğal dil işleme ve kredi riski değerlendirmesi gibi çeşitli uygulamalarda yaygın olarak kullanılmaktadır.
fit_forest = randomForest(Transported ~ ., data = train_train[, -1])
fit_forest$importance
## IncNodePurity
## HomePlanet 48.125593
## CryoSleep 107.989293
## Destination 30.575915
## Age 91.836683
## VIP 2.516955
## RoomService 125.936343
## FoodCourt 113.623386
## ShoppingMall 96.731658
## Spa 131.332461
## VRDeck 122.410460
## withgroup 12.925079
## deck 90.196804
## side 22.439063
## expense 238.242509
varImpPlot(fit_forest)
preds= predict(fit_forest, newdata = train_test[, -c(1,12)]) %>%
data.frame()
head(preds)
## .
## 1 0.1050685
## 2 0.1614667
## 3 0.8062586
## 4 0.1920834
## 5 0.6907569
## 6 0.1659998
Transported_pred_forest = ifelse(preds$. >0.5, 1, 0)
cm = table(Transported_train_test, Transported_pred_forest )
cm
## Transported_pred_forest
## Transported_train_test 0 1
## 0 823 256
## 1 175 919
(823+919)/(823+256+175+919)
## [1] 0.8016567
fit_forest = randomForest(Transported ~ ., data = train_c[, -1])
preds = predict(fit_forest, newdata = test_c) %>%
data.frame()
Transported_pred_forest = ifelse(preds$. > 0.5, TRUE, FALSE)
Transported = as.character(Transported_pred_forest)
PassengerId = test_c$PassengerId
Transported = as.vector(Transported)
sample_submission = cbind(PassengerId, Transported)
sample_submission = as.data.frame(sample_submission)
sample_submission$Transported = str_to_title(sample_submission$Transported)
write.csv(sample_submission, "sub_forest.csv", row.names = FALSE, quote = FALSE)