iv$.$ corr: Filter malicious transactions where Actual and Origin places are different. Calculate

elementwise product between Authentication$\_$score and IP$\_$validation$\_$score and then

perform correlation between the resultant vector and Amount column.

Output:

print$($corr$)=-0\mathrm{.}06$

v$.$ pca: Create a N x $5$ matrix where N is number of rows in the dataset and $5$ is the number of

columns, we will call these features $($Transaction$\_$type, Entry$\_$mode, Amount,

Authentication$\_$score, and IP$\_$validity$\_$score$)($before that you need to convert all

CITS $2401$

Computer Analysis

and Visualisation

Page $3$ of $6$

string values to numerical values. You can assume there will always be $3$ Transaction$\_$type and

use the following values $-$ ATM: $1,$ EFTPOS: $2,$ and Internet: $3$ and four Entry$\_$mode $-$ Magnetic

Stripe: $1,$ Manual: $2,$ Chip Card Read: $3,$ and NFC: $4).$ Calculate principal component analysis $($PCA$)$

to reduce the dimensionality of data to N X $1.$

The algorithm for PCA is:

a$)$ Standardize the data along all the features $($subtract mean and divide by standard deviation

over the feature dimension$).$

b$)$ Calculate the covariance matrix for the features

c$)$ Perform eigen decomposition on the covariance matrix to get eigenvectors $($principal

components$)$ and eigenvalues

d$)$ Sort the eigenvectors based on their eigenvalues from highest to lowest

e$)$ Select top k eigenvectors $($k$=1)$

f$)$ Transform the data using the selected eigenvectors $($dot product of eigenvectors and

Standardized data in step a$)$

Output:

print$($pca$.$shape$)=(10000,1)$

print$($pca$[0$:$5])=[0\mathrm{.}73,0\mathrm{.}81,0\mathrm{.}7,0\mathrm{.}93,0\mathrm{.}72]$