Credit Card Fraud Detection XGBOOST Guide
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 14
| Download | |
| Open PDF In Browser | View PDF |
Credit Card Fraud Detection - XGBOOST ∗
Javier Ng
EY
The datasets contains transactions made by credit cards in September 2013 by european cardholders. This dataset presents transactions that occurred in two days, where we have 492 frauds out
of 284,807 transactions. The dataset is highly unbalanced, the positive class (frauds) account for
0.172% of all transactions.
Keywords: xgboost, machine learning, R
Overview
Loading of Libraries
setwd("C:/Users/zm679xs/Desktop/R/Essential-R-Codes/EDA Projects/Credit Card Fraud Detection")
library(ggplot2) # Data visualization
library(readr) # CSV file I/O, e.g. the read_csv function
library(caret)
library(DMwR) # smote
library(xgboost)
library(Matrix)
library(reshape) #melt
library(pROC) # AUC
# Input data files are available in the "../input/" directory.
# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the
load("ccdata.RData")
Split Data
Split data into train, test and cv using caret.
## split into train, test, cv
set.seed(1900)
inTrain <- createDataPartition(y = ccdata$Class, p = .6, list = F) #60% train
train <- ccdata[inTrain,]
testcv <- ccdata[-inTrain,]
inTest <- createDataPartition(y = testcv$Class, p = .5, list = F) #20% test
test <- testcv[inTest,]
cv <- testcv[-inTest,]
train$Class <- as.factor(train$Class)
rm(inTrain, inTest, testcv) #20% cv
∗ PDF is available on EY FDA shared drive. Current version: March 11, 2019; Corresponding author: javier.ng@sg.
ey.com.
1
Imbalance Dataset
Using SMOTE to make the dataset more balanced.
#using SMOTE instead
i <- grep("Class", colnames(train)) # Get index Class column
train_smote <- SMOTE(Class ~ ., as.data.frame(train), perc.over = 20000, perc.under=100)
table(train_smote$Class)
##
##
0
1
## 60400 60702
#Identify the Predictors and the dependent variable, aka label.
predictors <- colnames(train_smote[-ncol(train_smote)]) #remove last column
#xgboost works only if the labels are numeric. Hence, convert the labels (Species) to numeric.
label <- as.numeric(train_smote[,ncol(train_smote)])
#Alas, xgboost works only if the numeric labels start from 0. Hence, subtract 1 from the label.
label <- as.numeric(train_smote[,ncol(train_smote)])-1
print(table(label))
## label
##
0
1
## 60400 60702
Setting Parameters for XGBoost
# set parameters:
parameters <- list(
# General Parameters
booster
=
silent
=
# Booster Parameters
eta
=
gamma
=
max_depth
=
min_child_weight =
subsample
=
colsample_bytree =
colsample_bylevel =
lambda
=
alpha
=
# Task Parameters
objective
=
eval_metric
=
seed
=
)
"gbtree",
0,
0.3,
0,
6,
1,
1,
1,
1,
1,
0,
"binary:logistic",
"auc",
1900
2
Train Model
Train the model with the parameters set above and nrounds = 25 (increasing nrounds does not
improve the model anymore). Plots show increasing train and cv AUC in the beginning and
stagnating at later rounds as expected.
## Training of models
# Original
cv.nround = 200; # Number of rounds. This can be set to a lower or higher value, if you wish, e
bst.cv <- xgb.cv(
param=parameters,
data = as.matrix(train_smote[,predictors]),
label = label,
nfold = 3,
nrounds=cv.nround,
prediction=T)
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
train-auc:0.988126+0.000290
train-auc:0.994458+0.000626
train-auc:0.996438+0.000303
train-auc:0.997409+0.000210
train-auc:0.998177+0.000100
train-auc:0.998640+0.000145
train-auc:0.998912+0.000079
train-auc:0.999116+0.000082
train-auc:0.999324+0.000044
train-auc:0.999437+0.000048
train-auc:0.999559+0.000040
train-auc:0.999609+0.000043
train-auc:0.999674+0.000033
train-auc:0.999721+0.000038
train-auc:0.999774+0.000020
train-auc:0.999816+0.000006
train-auc:0.999855+0.000010
train-auc:0.999868+0.000009
train-auc:0.999893+0.000004
train-auc:0.999911+0.000008
train-auc:0.999932+0.000011
train-auc:0.999945+0.000008
train-auc:0.999952+0.000010
train-auc:0.999959+0.000010
train-auc:0.999964+0.000008
train-auc:0.999969+0.000007
train-auc:0.999976+0.000005
train-auc:0.999979+0.000005
train-auc:0.999982+0.000004
train-auc:0.999983+0.000004
train-auc:0.999985+0.000004
test-auc:0.987120+0.000203
test-auc:0.993682+0.000973
test-auc:0.995692+0.000592
test-auc:0.996878+0.000309
test-auc:0.997759+0.000171
test-auc:0.998255+0.000200
test-auc:0.998543+0.000171
test-auc:0.998800+0.000144
test-auc:0.999043+0.000009
test-auc:0.999184+0.000020
test-auc:0.999334+0.000016
test-auc:0.999400+0.000026
test-auc:0.999473+0.000032
test-auc:0.999529+0.000021
test-auc:0.999595+0.000016
test-auc:0.999658+0.000022
test-auc:0.999710+0.000013
test-auc:0.999729+0.000014
test-auc:0.999765+0.000009
test-auc:0.999791+0.000008
test-auc:0.999824+0.000011
test-auc:0.999845+0.000012
test-auc:0.999855+0.000020
test-auc:0.999868+0.000027
test-auc:0.999874+0.000027
test-auc:0.999884+0.000025
test-auc:0.999892+0.000026
test-auc:0.999909+0.000024
test-auc:0.999918+0.000021
test-auc:0.999922+0.000019
test-auc:0.999926+0.000021
3
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
train-auc:0.999987+0.000003
train-auc:0.999988+0.000001
train-auc:0.999990+0.000002
train-auc:0.999990+0.000002
train-auc:0.999991+0.000003
train-auc:0.999993+0.000002
train-auc:0.999993+0.000002
train-auc:0.999994+0.000002
train-auc:0.999996+0.000002
train-auc:0.999997+0.000002
train-auc:0.999997+0.000002
train-auc:0.999998+0.000002
train-auc:0.999998+0.000001
train-auc:0.999998+0.000001
train-auc:0.999999+0.000001
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
test-auc:0.999931+0.000020
test-auc:0.999940+0.000017
test-auc:0.999942+0.000018
test-auc:0.999946+0.000021
test-auc:0.999949+0.000021
test-auc:0.999953+0.000022
test-auc:0.999955+0.000021
test-auc:0.999959+0.000020
test-auc:0.999962+0.000019
test-auc:0.999964+0.000019
test-auc:0.999966+0.000018
test-auc:0.999967+0.000017
test-auc:0.999970+0.000017
test-auc:0.999971+0.000018
test-auc:0.999972+0.000017
test-auc:0.999973+0.000018
test-auc:0.999974+0.000017
test-auc:0.999975+0.000016
test-auc:0.999977+0.000014
test-auc:0.999978+0.000014
test-auc:0.999979+0.000013
test-auc:0.999979+0.000014
test-auc:0.999979+0.000015
test-auc:0.999981+0.000013
test-auc:0.999981+0.000013
test-auc:0.999981+0.000013
test-auc:0.999983+0.000011
test-auc:0.999982+0.000013
test-auc:0.999983+0.000012
test-auc:0.999984+0.000011
test-auc:0.999985+0.000010
test-auc:0.999985+0.000010
test-auc:0.999984+0.000011
test-auc:0.999985+0.000010
test-auc:0.999985+0.000010
test-auc:0.999985+0.000011
test-auc:0.999986+0.000010
test-auc:0.999987+0.000010
test-auc:0.999987+0.000010
test-auc:0.999987+0.000010
test-auc:0.999987+0.000009
test-auc:0.999987+0.000009
test-auc:0.999987+0.000009
test-auc:0.999987+0.000009
test-auc:0.999988+0.000009
test-auc:0.999987+0.000009
test-auc:0.999988+0.000009
test-auc:0.999988+0.000009
4
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[80] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[81] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[82] train-auc:1.000000+0.000000 test-auc:0.999988+0.000009
[83] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[84] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[85] train-auc:1.000000+0.000000 test-auc:0.999988+0.000009
[86] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[87] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[88] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[89] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[90] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[91] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[92] train-auc:1.000000+0.000000 test-auc:0.999987+0.000009
[93] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[94] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[95] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[96] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[97] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[98] train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[99] train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[100]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[101]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[102]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[103]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[104]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[105]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[106]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[107]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[108]
train-auc:1.000000+0.000000 test-auc:0.999986+0.000011
[109]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[110]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[111]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[112]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[113]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[114]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[115]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[116]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000011
[117]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[118]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[119]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[120]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[121]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[122]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[123]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[124]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[125]
train-auc:1.000000+0.000000 test-auc:0.999988+0.000010
[126]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000010
[127]
train-auc:1.000000+0.000000 test-auc:0.999987+0.000011
5
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[128]
[129]
[130]
[131]
[132]
[133]
[134]
[135]
[136]
[137]
[138]
[139]
[140]
[141]
[142]
[143]
[144]
[145]
[146]
[147]
[148]
[149]
[150]
[151]
[152]
[153]
[154]
[155]
[156]
[157]
[158]
[159]
[160]
[161]
[162]
[163]
[164]
[165]
[166]
[167]
[168]
[169]
[170]
[171]
[172]
[173]
[174]
[175]
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
test-auc:0.999987+0.000011
test-auc:0.999987+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000009
test-auc:0.999988+0.000009
test-auc:0.999988+0.000009
test-auc:0.999988+0.000009
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999988+0.000010
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
6
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[176]
[177]
[178]
[179]
[180]
[181]
[182]
[183]
[184]
[185]
[186]
[187]
[188]
[189]
[190]
[191]
[192]
[193]
[194]
[195]
[196]
[197]
[198]
[199]
[200]
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
train-auc:1.000000+0.000000
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999988+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
test-auc:0.999987+0.000011
#Find where the minimum logloss occurred
min.loss.idx = which.max(bst.cv$evaluation_log[, test_auc_mean])
cat("Minimum logloss occurred in round : ", min.loss.idx, "\n")
## Minimum logloss occurred in round :
139
# Minimum logloss
print(bst.cv$evaluation_log[min.loss.idx,])
##
iter train_auc_mean train_auc_std test_auc_mean test_auc_std
## 1: 139
1
0
0.9999883 9.392662e-06
Predict
Predict with dataset and set threshold of 0.5
## Make predictions
bst <- xgboost(
param=parameters,
data =as.matrix(train_smote[,predictors]), #training it without the output variable!
label = label,
nrounds=min.loss.idx)
7
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
train-auc:0.988286
train-auc:0.995025
train-auc:0.996358
train-auc:0.997692
train-auc:0.998161
train-auc:0.998605
train-auc:0.999029
train-auc:0.999174
train-auc:0.999332
train-auc:0.999449
train-auc:0.999527
train-auc:0.999583
train-auc:0.999652
train-auc:0.999680
train-auc:0.999759
train-auc:0.999798
train-auc:0.999830
train-auc:0.999855
train-auc:0.999878
train-auc:0.999900
train-auc:0.999931
train-auc:0.999945
train-auc:0.999957
train-auc:0.999963
train-auc:0.999971
train-auc:0.999975
train-auc:0.999977
train-auc:0.999980
train-auc:0.999982
train-auc:0.999985
train-auc:0.999987
train-auc:0.999989
train-auc:0.999989
train-auc:0.999990
train-auc:0.999991
train-auc:0.999993
train-auc:0.999994
train-auc:0.999995
train-auc:0.999995
train-auc:0.999996
train-auc:0.999996
train-auc:0.999997
train-auc:0.999998
train-auc:0.999998
train-auc:0.999999
train-auc:0.999999
train-auc:0.999999
train-auc:0.999999
8
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
[85]
[86]
[87]
[88]
[89]
[90]
[91]
[92]
[93]
[94]
[95]
[96]
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
train-auc:1.000000
9
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
[97] train-auc:1.000000
[98] train-auc:1.000000
[99] train-auc:1.000000
[100]
train-auc:1.000000
[101]
train-auc:1.000000
[102]
train-auc:1.000000
[103]
train-auc:1.000000
[104]
train-auc:1.000000
[105]
train-auc:1.000000
[106]
train-auc:1.000000
[107]
train-auc:1.000000
[108]
train-auc:1.000000
[109]
train-auc:1.000000
[110]
train-auc:1.000000
[111]
train-auc:1.000000
[112]
train-auc:1.000000
[113]
train-auc:1.000000
[114]
train-auc:1.000000
[115]
train-auc:1.000000
[116]
train-auc:1.000000
[117]
train-auc:1.000000
[118]
train-auc:1.000000
[119]
train-auc:1.000000
[120]
train-auc:1.000000
[121]
train-auc:1.000000
[122]
train-auc:1.000000
[123]
train-auc:1.000000
[124]
train-auc:1.000000
[125]
train-auc:1.000000
[126]
train-auc:1.000000
[127]
train-auc:1.000000
[128]
train-auc:1.000000
[129]
train-auc:1.000000
[130]
train-auc:1.000000
[131]
train-auc:1.000000
[132]
train-auc:1.000000
[133]
train-auc:1.000000
[134]
train-auc:1.000000
[135]
train-auc:1.000000
[136]
train-auc:1.000000
[137]
train-auc:1.000000
[138]
train-auc:1.000000
[139]
train-auc:1.000000
# Make prediction on the testing data.
test$prediction <- predict(bst, as.matrix(test[,predictors])) #here, I've removed the output var
10
test$prediction <- ifelse(test$prediction >= 0.5, 1 , 0)
Confusion Matrix
Sensitivity (TPR) = 0.9991 Specificity (TNR) = 0.9135 Accuracy = 0.9989
#Compute the accuracy of predictions.
confmatrix_table <- confusionMatrix(as.factor(test$prediction), as.factor(test$Class)) #sensitiv
#accuracy = 0.9989
plot_confusion_matrix <- function(test_df, sSubtitle) {
tst <- data.frame(round(test_df$prediction,0), test_df$Class)
opts <- c("Predicted", "True")
names(tst) <- opts
cf <- plyr::count(tst)
cf[opts][cf[opts]==0] <- "Not Fraud"
cf[opts][cf[opts]==1] <- "Fraud"
ggplot(data = cf, mapping = aes(x = True, y = Predicted)) +
labs(title = "Confusion matrix", subtitle = sSubtitle) +
geom_tile(aes(fill = freq), colour = "grey") +
geom_text(aes(label = sprintf("%1.0f", freq)), vjust = 1) +
scale_fill_gradient(low = "lightblue", high = "Green") +
theme_bw() + theme(legend.position = "none")
}
plot_confusion_matrix(test, paste("XGBoost with", paste("min logloss at round: ", min.loss.idx,
"Sensitivity:", round(confmatrix_table[[4]][1], 4), "\n",
"Specificity:", round(confmatrix_table[[4]][2], 4)))
11
Confusion matrix
XGBoost with min logloss at round: 139
Sensitivity: 0.9991
Specificity: 0.9135
9
56803
95
54
Predicted
Not Fraud
Fraud
Fraud
Not Fraud
True
Importance Matrix
Identify features that are most important
##plot feature importance
importance_matrix <- xgb.importance(model = bst)
xgb.plot.importance(importance_matrix)
12
V14
V4
V8
V11
V10
V12
V1
V26
V3
V17
Amount
V7
V13
V18
Time
V23
V19
V2
V15
V20
V21
V9
V5
V25
V6
V24
V22
V27
V16
V28
0.0
0.1
0.2
0.3
0.4
ROC Curve
AUC = 0.956
###############################################################
library(ROCR)
## Loading required package: gplots
##
## Attaching package: 'gplots'
## The following object is masked from 'package:stats':
##
##
lowess
# Use ROCR package to plot ROC Curve
xgb.pred <- prediction(test$prediction, test$Class)
xgb.perf <- performance(xgb.pred, "tpr", "fpr")
plot(xgb.perf,
avg="threshold",
colorize=TRUE,
13
0.5
0.6
lwd=1,
main="ROC Curve w/ Thresholds",
print.cutoffs.at=seq(0, 1, by=0.05),
text.adj=c(-0.5, 0.5),
text.cex=0.5)
grid(col="lightgray")
axis(1, at=seq(0, 1, by=0.1))
axis(2, at=seq(0, 1, by=0.1))
abline(v=c(0.1, 0.3, 0.5, 0.7, 0.9), col="lightgray", lty="dotted")
abline(h=c(0.1, 0.3, 0.5, 0.7, 0.9), col="lightgray", lty="dotted")
lines(x=c(0, 1), y=c(0, 1), col="black", lty="dotted")
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.2
0.8
0.4
0.2
0.4
0.6
1.6
0.2
0.8
0.1
0.05 0
0
0.0
Average true positive rate
1
0.55 0.5
0.15 0.1
0.65 0.6
0.25 0.2
0.75 0.7
0.35 0.3
0.85 0.8
0.45 0.4
0.95 0.9
2
1.0
ROC Curve w/ Thresholds
0.0
Average false positive rate
auc_ROCR <- performance(xgb.pred, measure = "auc") #gives the AUC rate
auc_ROCR <- auc_ROCR@y.values[[1]]
auc_ROCR
## [1] 0.9562559
#AUC = 0.956
14
1.0
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.5 Linearized : No Page Count : 14 Page Mode : UseOutlines Author : Javier Ng (EY) Title : Credit Card Fraud Detection - XGBOOST Subject : Creator : LaTeX with hyperref Producer : pdfTeX-1.40.20 Keywords : xgboost, machine learning, R Create Date : 2019:03:11 13:34:16+08:00 Modify Date : 2019:03:11 13:34:16+08:00 Trapped : False PTEX Fullbanner : This is MiKTeX-pdfTeX 2.9.6959 (1.40.20)EXIF Metadata provided by EXIF.tools