R Notebook

Load the Library

xgboost <- xgb.train is an advanced interface for training an xgboost model. The xgboost function is a simpler wrapper for xgb.train.

magrittr <- The operators pipe their left-hand side values forward into expressions that appear on the right-hand side, i.e. one can replace f(x) with x %>% f, where %>% is the (main) pipe-operator.

dplyr <- %>%

Matrix <- Construct a Matrix of a class that inherits from Matrix

library(xgboost)

Warning: package ‘xgboost’ was built under R version 4.1.3
Registered S3 method overwritten by 'data.table':
  method           from
  print.data.table     

Attaching package: ‘xgboost’

The following object is masked from ‘package:dplyr’:

    slice

read the data

source: https://raw.githubusercontent.com/bkrai/Top-10-Machine-Learning-Methods-With-R/master/binary.csv

we need to change the rank into factor not “integer”

class(data$rank)

[1] "integer"

data$rank <- as.factor(data$rank)

we can use str() to check data type

str(data$rank)

 Factor w/ 4 levels "1","2","3","4": 3 3 1 4 4 2 1 2 3 2 ...

Making model Train and Test

Splitting data 80/20

set.seed(123) # same observation 
ind <- sample(2, nrow(data), replace = T, prob = c(0.8, 0.2))
train <- data[ind==1,]
test <- data[ind==2,]

train model

train_label <- train[,"admit"]
train_matrix <- xgb.DMatrix(data = as.matrix(trainm), label = train_label)

test model

# test model
testm <- sparse.model.matrix(admit~.-1, data = test)
test_label <- test[,"admit"]
test_matrix <- xgb.DMatrix(data = as.matrix(testm), label= test_label)

xgb_params <- list("objective" = "multi:softprob",
                   "eval_metric" = "mlogloss",
                   "num_class" = nc)

watchlist <- list(train = train_matrix, test = test_matrix)

# eXtreme Gradian Boosting Model
bst_model <- xgb.train(params = xgb_params,
                       data = train_matrix,
                       nrounds = 107,
                       watchlist = watchlist,
                       eta = 0.001, # more robust to overfitting
                       max.depth = 3,
                       gamma = 0,
                       subsample = 1,
                       colsample_bytree = 1,
                       missing = NA,
                       seed = 333)

# Training and test error plot
e <- data.frame(bst_model$evaluation_log)
plot(e$iter, e$train_mlogloss, col = 'blue')
lines(e$iter,e$test_mlogloss, col = 'red')
min(e$test_mlogloss)
# not the best model because of the red line

# e[e$test_mlogloss == 0.6014728,]

imp <- xgb.importance(colnames(train_matrix), model = bst_model)
print(imp)
xgb.plot.importance(imp)

# Prediction and confusin matrix - test data
p <- predict(bst_model, newdata = test_matrix)

pred <- matrix(p, nrow = nc, ncol = length(p)/nc) %>% 
  t() %>% 
  data.frame() %>% 
  mutate(label = test_label, max_prob = max.col(.,"last")-1)
head(pred)

Confussion Matrix Result (test)

59 out of 75 prediction is correct from data test 47+3 78% accurate

table(Prediction = pred$max_prob, Actual = pred$label)
59/75