Machine Learning & Big Data Project

Credit Card Fraud Detection: Data Pre-processing & Visualisation

Abstract

The Data pre-processing and visualization herein is done with the help of R Programming Language and Tableau. First, I will process non dummy variable in Section 1 and then process dummy variables in Section 2

Tableau Link to my Profile for interactive Visualization: https://public.tableau.com/app/profile/emmanuel.m.maseruka

Section 1: Non-Dummy Variable Preprocessing

distancefromhome - the distance from home where the transaction happened.

distancefromlast_transaction - the distance from last transaction happened.

ratiotomedianpurchaseprice - Ratio of purchased price transaction to median purchase price.

1) STEP 1

Pulling my Credit card Fraud Data set

data<-read.csv("C:\\Users\\Kevin Meng\\OneDrive\\Desktop\\1.Credit card Fraud.csv",header = TRUE)%>%as.data.frame() 

2) STEP 2

Normalization of Non Dummy Attributes

range<-max(data$distance_from_home)-min(data$distance_from_home)

dist.home.norm<-(data$distance_from_home-min(data$distance_from_home))/range   #normalized distance from home

range<-max(data$distance_from_last_transaction)-min(data$distance_from_last_transaction)

dist.transac.norm<-(data$distance_from_last_transaction-min(data$distance_from_last_transaction))/range  #normalized distance from last transaction

range<-max(data$ratio_to_median_purchase_price)-min(data$ratio_to_median_purchase_price)

median2purchase.ratio.norm<-(data$ratio_to_median_purchase_price-min(data$ratio_to_median_purchase_price))/range #normalized ratio_to_median_purchase_price

3) STEP 3

New Dataset with Normalized Attributes

new.data<-cbind(dist.home.norm,dist.transac.norm,median2purchase.ratio.norm,data$repeat_retailer,data$used_chip,data$used_pin_number,data$online_order,data$fraud) #Creating new dataset with normalised attributes

4) STEP 4

Visualizing the Relationship between the Normalized Attributes

The Visualization below was produced by Tableau

Tableau Link to my Profile for interactive Visualistion: https://public.tableau.com/app/profile/emmanuel.m.maseruka/viz/CreditCardFraud_16558266779080/Dashboard1

Section 2: Dummy Variable Preprocessing & Visualisation

repeat_retailer - Is the transaction happened from same retailer. (dummy)

used_chip - Is the transaction through chip (dummy)

used_pin_number - Is the transaction happened by using PIN number(dummy)

online_order - Is the transaction an online order (dummy)

1) STEP 1

Subdividing Attributes into different groups

repeat_retailer.fraud1<-subset(data,fraud==1, select = repeat_retailer)
repeat_retailer.1.fraud1<-subset(repeat_retailer.fraud1,repeat_retailer==1)       #Repeat Retailers that were a fraud
repeat_retailer.0.fraud1<-subset(repeat_retailer.fraud1,repeat_retailer==0)       #Non Repeat Retailers that were a fraud




repeat_retailer.fraud0<-subset(data,fraud==0, select = repeat_retailer)
repeat_retailer.1.fraud0<-subset(repeat_retailer.fraud0,repeat_retailer==1)        #Repeat Retailers that were not fraud
repeat_retailer.0.fraud0<-subset(repeat_retailer.fraud0,repeat_retailer==0)        #Non Repeat Retailers that were not fraud

fraud<-c(nrow(repeat_retailer.1.fraud1),nrow(repeat_retailer.0.fraud1),nrow(repeat_retailer.fraud1))
no.fraud<-c(nrow(repeat_retailer.1.fraud0),nrow(repeat_retailer.0.fraud0),nrow(repeat_retailer.fraud0))



totals<-c(nrow(repeat_retailer.1.fraud1)+nrow(repeat_retailer.1.fraud0),nrow(repeat_retailer.0.fraud1)+nrow(repeat_retailer.0.fraud0),nrow(repeat_retailer.fraud1)+nrow(repeat_retailer.fraud0))



x<-rbind(fraud,no.fraud,totals) 
x<-as.data.frame(x)
colnames(x)<-c("Repeat Retailers","Non Repeat Retailers", "Total")
row.names(x)<-c("Fraud","No Fraud","Total")
x %>%
  kbl() %>%
  kable_material(c("striped", "hover"))

	Repeat Retailers	Non Repeat Retailers	Total
Fraud	76925	10478	87403
No Fraud	804611	107986	912597
Total	881536	118464	1000000

NA

used_chip.fraud1<-subset(data,fraud==1, select = used_chip)
used_chip.1.fraud1<-subset(used_chip.fraud1,used_chip==1)                         #Instances who Used_chip that were a fraud
used_chip.0.fraud1<-subset(used_chip.fraud1,used_chip==0)                         #Instances who did not Use_chip that were a fraud


used_chip.fraud0<-subset(data,fraud==0, select = used_chip)
used_chip.1.fraud0<-subset(used_chip.fraud0,used_chip==1)                         #Instances who Used_chip that were a fraud
used_chip.0.fraud0<-subset(used_chip.fraud0,used_chip==0)                         #Instances who Used_chip that were a fraud

fraud2<-c(nrow(used_chip.1.fraud1),nrow(used_chip.0.fraud1),nrow(used_chip.fraud1))
no.fraud2<-c(nrow(used_chip.1.fraud0),nrow(used_chip.0.fraud0),nrow(used_chip.fraud0))
y<-rbind(fraud2,no.fraud2)


totals1<-c(nrow(used_chip.1.fraud1)+nrow(used_chip.1.fraud0),nrow(used_chip.0.fraud1)+nrow(used_chip.0.fraud0),nrow(used_chip.fraud1)+nrow(used_chip.fraud0))



y<-rbind(fraud2,no.fraud2,totals1) 
y<-as.data.frame(y)
colnames(y)<-c("Used Chip","No Chip", "Total")
row.names(y)<-c("Fraud","No Fraud","Total")
y %>%
  kbl() %>%
  kable_material(c("striped", "hover"))

	Used Chip	No Chip	Total
Fraud	22410	64993	87403
No Fraud	327989	584608	912597
Total	350399	649601	1000000

used_pin_number.fraud1<-subset(data,fraud==1, select = used_pin_number)
used_pin_number.1.fraud1<-subset(used_pin_number.fraud1,used_pin_number==1)        #Instances who Used_pin that were a fraud
used_pin_number.0.fraud1<-subset(used_pin_number.fraud1,used_pin_number==0)        #Instances who did not Use_chip that were a fraud


used_pin_number.fraud0<-subset(data,fraud==0, select = used_pin_number)
used_pin_number.1.fraud0<-subset(used_pin_number.fraud0,used_pin_number==1)        #Instances who Used_chip that were not a fraud
used_pin_number.0.fraud0<-subset(used_pin_number.fraud0,used_pin_number==0)        #Instances who did not Use_chip that were not a fraud

fraud3<-c(nrow(used_pin_number.1.fraud1),nrow(used_pin_number.0.fraud1),nrow(used_pin_number.fraud1))
no.fraud3<-c(nrow(used_pin_number.1.fraud0),nrow(used_pin_number.0.fraud0),nrow(used_pin_number.fraud0))
z<-rbind(fraud3,no.fraud3)



totals2<-c(nrow(used_pin_number.1.fraud1)+nrow(used_pin_number.1.fraud0),nrow(used_pin_number.0.fraud1)+nrow(used_pin_number.0.fraud0),nrow(used_pin_number.fraud1)+nrow(used_pin_number.fraud0))



z<-rbind(fraud3,no.fraud3,totals2) 
z<-as.data.frame(z)
colnames(z)<-c("Used Pin","No Pin", "Total")
row.names(z)<-c("Fraud","No Fraud","Total")
z %>%
  kbl() %>%
  kable_material(c("striped", "hover"))

	Used Pin	No Pin	Total
Fraud	273	87130	87403
No Fraud	100335	812262	912597
Total	100608	899392	1000000

online_order.fraud1<-subset(data,fraud==1, select = online_order)
online_order.1.fraud1<-subset(online_order.fraud1,online_order==1)                 #Instances who Used_online purchases that were a fraud
online_order.0.fraud1<-subset(online_order.fraud1,online_order==0)                 #Instances who did not Use_online purchases that were a fraud

online_order.fraud0<-subset(data,fraud==0, select = online_order)
online_order.1.fraud0<-subset(online_order.fraud0,online_order==1)                 #Instances who Used_online purchases that were not a fraud
online_order.0.fraud0<-subset(online_order.fraud0,online_order==0)                #Instances who did not Use_online purchases that were not a fraud


fraud4<-c(nrow(online_order.1.fraud1),nrow(online_order.0.fraud1),nrow(online_order.fraud1))
no.fraud4<-c(nrow(online_order.1.fraud0),nrow(online_order.0.fraud0),nrow(online_order.fraud0))
a<-rbind(fraud4,no.fraud4)


totals3<-c(nrow(online_order.1.fraud1)+nrow(online_order.1.fraud0),nrow(online_order.0.fraud1)+nrow(online_order.0.fraud0),nrow(online_order.fraud1)+nrow(online_order.fraud0))



a<-rbind(fraud4,no.fraud4,totals3) 
a<-as.data.frame(a)
colnames(a)<-c("Online Order","Offline Order", "Total")
row.names(a)<-c("Fraud","No Fraud","Total")
a %>%
  kbl() %>%
  kable_material(c("striped", "hover"))

	Online Order	Offline Order	Total
Fraud	82711	4692	87403
No Fraud	567841	344756	912597
Total	650552	349448	1000000

2) STEP 2

Visualizing the Dummy Variables

The Visualization below was produced by Tableau

Tableau Link to my Profile for interactive Visualistion: https://public.tableau.com/app/profile/emmanuel.m.maseruka/viz/CreditCardFraudDetectionVisualisationofDummyVariables2/Dashboard1

Tableau Link to my Profile for interactive Visualiztion: https://public.tableau.com/app/profile/emmanuel.m.maseruka/viz/CreditCardFraudDetectionVisualisationofDummyVariables1/Dashboard2