Make a rmarkdown in this code: library$($tidyverse$)$ library$($caret$)$ library$($cluster$)$ library$($factoextra$)$ # Load the dataset data $<-$ read.csv$("$C:$/$Users$/$arthu$/$Downloads$/$bank$\_$transactions$\_$data$\_2.$csv$")$ # $2.$ Explore Your Data $5$ Points ## a$.$ Basic Exploration # View summary statistics of the dataset cat$("$Summary of the dataset:

$")$ summary$($data$)$ # Check for missing values cat$("$

Missing values:

$")$ print$($colSums$($is$.$na$($data$)))$ # Histograms to view the distribution of key variables ggplot$($data$,$ aes$($x $=$ TransactionAmount$))+$ geom$\_$histogram$($bins $=30,$ fill $=$ "blue", color $=$ "black"$)+$ labs$($title $=$ "Distribution of Transaction Amount"$)$ ggplot$($data$,$ aes$($x $=$ AccountBalance$))+$ geom$\_$histogram$($bins $=30,$ fill $=$ "green", color $=$ "black"$)+$ labs$($title $=$ "Distribution of Account Balance"$)$ # Scatterplot to explore relationships between variables ggplot$($data$,$ aes$($x $=$ TransactionAmount, y $=$ AccountBalance$))+$ geom$\_$point$($alpha $=0\mathrm{.}7)+$ labs$($title $=$ "Scatterplot: Transaction Amount vs Account Balance"$)$ # Outlier detection using IQR method for TransactionAmount Q$1<-$ quantile$($data$TransactionAmount, $0\mathrm{.}25)$ Q$3<-$ quantile$($data$TransactionAmount, $0\mathrm{.}75)$ IQR $<-$ Q$3-$ Q$1$ outliers $<-$ data $|>$ filter$($TransactionAmount $<($Q$1-1\mathrm{.}5*$ IQR$)|$ TransactionAmount $>($Q$3+1\mathrm{.}5*$ IQR$))$ cat$("$

Outliers detected for TransactionAmount:

$")$ print$($outliers$)$ # Normalize the dataset numeric$\_$cols $<-$ sapply$($data$,$ is$.$numeric$)$ numeric$\_$data $<-$ data $|>$ select$($where$($is$.$numeric$))$ numeric$\_$scaled $<-$ scale$($numeric$\_$data$)$ # Step $3$: # Handle missing values numeric$\_$cols $<-$ sapply$($data$,$ is$.$numeric$)$ categorical$\_$cols $<-$ sapply$($data$,$ is$.$character$)$ # Numeric columns with median data$[$numeric$\_$cols$]<-$ lapply$($data$[$numeric$\_$cols$],$ function$($x$)$ ifelse$($is$.$na$($x$),$ median$($x$,$ na$.$rm $=$ TRUE$),$ x$))$ # Fill categorical columns with mode for $($col in names$($data$)[$categorical$\_$cols$])\{$ mode$\_$value $<-$ as$.$character$($sort$($table$($data$[[$col$]]),$ decreasing $=$ TRUE$)[1])$ data$[[$col$]]<-$ ifelse$($is$.$na$($data$[[$col$]]),$ mode$\_$value, data$[[$col$]])\}$ # Step $4$: # Objective: Detect fraudulent transaction patterns using clustering numeric$\_$scaled $<-$ scale$($data $|>$ select$($where$($is$.$numeric$)))$ # Perform K$-$Means clustering set.seed$(42)$ kmeans$\_$result $<-$ kmeans$($numeric$\_$scaled, centers $=5,$ nstart $=25)$ data$KMeans$\_$Cluster $<-$ kmeans$\_$result$cluster # Calculate distances from centroids. # Centroids refer to the central points of clusters in a clustering algorithm. centroids $<-$ kmeans$\_$result$centers distances $<-$ apply$($numeric$\_$scaled, $1,$ function$($x$)$ min$($sqrt$($colSums$(($t$($centroids$)-$ x$)^2))))$ data$KMeans$\_$Distance $<-$ distances # Mean $+3*$ Standard Deviation: Identifying Fraud # The threshold for identifying potential fraud is set by calculating the mean distance from the centroids $($the center of the clusters$)$ # and adding $3$ times the standard deviation of the distances. This threshold is commonly used in anomaly detection, as points that # fall further than $3$ standard deviations from the mean are considered outliers $($fraudulent transactions, in this case$).$ threshold $<-$ mean$($distances$)+3*$ sd$($distances$)$ # Flagging Fraudulent Transactions: # The 'distances' variable contains the distance of each point $($transaction$)$ from its cluster centroid. # Any transaction with a distance greater than the threshold is flagged as potentially fraudulent. # The result is stored in the new column 'KMeans$\_$Fraud', where TRUE indicates a fraud, and FALSE indicates a normal transaction. data$KMeans$\_$Fraud $<-$ distances $>$ threshold # Summary fraud$\_$summary $<-$ table$($data$KMeans$\_$Fraud$)$ cat$("$

Fraud Summary:

$")$ print$($fraud$\_$summary$)$ # The objective of this part of the code is to visualize how the data points have been grouped into clusters based on # the features TransactionAmount and AccountBalance, which were used in the #K$-$Means clustering algorithm. This helps in understanding how well the clustering algorithm has separated the data. # Visualize K$-$Means Clusters with two key features: TransactionAmount and AccountBalance ggplot$($data$,$ aes$($x $=$ TransactionAmount, y $=$ AccountBalance, color $=$ as$.$factor$($KMeans$\_$Cluster$)))+$ geom$\_$point$($alpha $=0\mathrm{.}7)+$ labs$($title $="$K$-$Means clustering algorithm", x $=$ "Transaction Amount", y $=$ "Account Balance"$)+$ scale$\_$color$\_$viridis$\_$d$()+$ theme$\_$minimal$()$ # Highlight fraud points # Red points are fraud detected fraud$\_$points $<-$ data $|>$ filter$($KMeans$\_$Fraud $==$ TRUE$)$ ggplot$($data$,$ aes$($x $=$ TransactionAmount, y $=$ AccountBalance, color $=$ as$.$factor$($KMeans$\_$Cluster$)))+$ geom$\_$point$($alpha $=0\mathrm{.}7)+$ geom$\_$point$($data $=$ fraud$\_$points, aes$($x $=$ TransactionAmount, y $=$ AccountBalance$),$ color $=$ "red", size $=3)+$ labs$($title $="$K$-$Means Clusters with Fraud Points", x $=$ "Transaction Amount", y $=$ "Account Balance"$)+$ scale$\_$color$\_$viridis$\_$d$()+$ theme$\_$minimal$()$ # Fraud detection logic is based on distance from centroids cat$("$

Total Fraudulent Transactions Detected $($Using K$-$Means clustering$)$:$",$ nrow$($fraud$\_$points$),"$

$")$ cat$("$

Fraudulent Transactions Detected:

$")$ print$($fraud$\_$points$)$