Assignment 2 - Video Presentation (35%)

Principal Component Analysis

TASK

For your video presentation, you must demonstrate your PCA analysis on the continuous features of the WACY-COM dataset and interpret the results. Submit the recording via the Panopto link on Canvas. Please ensure you follow the instructions  carefully.

The due date for this assessment is Friday of Week 6 on 4 April 2025 before midnight.

Perform PCA and Visualise Data

(i) First, copy the code below to a R script. Enter your student ID into the command set.seed(.) and run the whole code. The code will create a sub-sample of 400 that is unique to you.

#You may need to change/include the path of your working directory

#Import the dataset into R Studio.

dat <- read.csv("WACY-COM.csv", na.strings=NA, stringsAsFactors=TRUE) set.seed(Enter your student ID here)

#Randomly select 400 rows

selected.rows <- sample(1:nrow(dat),size=400,replace=FALSE)

#Your sub-sample of 400 observations

mydata <- dat[selected.rows,]

dim(mydata) #check the dimension of your sub-sample

(ii)  Extract  only  the continuous features and  the APT feature from  the  WACY-COM  dataset and store them as a data frame/tibble. Refer to Assignment 1 for the feature description if needed.

(iii) Clean the extracted data based on the feedback received from Assignment 1.

(iv) Remove the incomplete cases to make it usable in “R” for PCA.

(v)  Perform PCA using prcomp(.) in R, but only on the numeric features (i.e. ignore APT  in this step).

-     Explain   why   you   believe   the   data   should   or   should   not   be   scaled,   i.e. standardised, when performing PCA.

-     Display and describe the individual and cumulative proportions of variance (3 decimal places) explained by each of the principal components.

-     Outline how many principal components are adequate to explain at least 50% of the variability in your data.

-     Display and interpret the coefficients (or loadings) to 3 decimal places for PC1, PC2 and  PC3.  Describe  which  features  (based  on  the  loadings)  are  the key drivers for each of these three principal components.

(vi) Create and display the biplot for PC1 vs. PC2 to visualise the PCA results in the first two dimensions. Colour-code the points based on the APT feature. Explain the biplot by commenting on the PCA plot and the loadings plot individually, and then both plots combined (see Slides 28-29 of Module 3 notes). Finally, comment on and justify which (if any) features can help distinguish APT activity.

(vii)     Based on the results from parts (v) and (vi), describe

- whether PC1 or PC2 (choose one) best assists in classifying APT. Hint: Project   all points in the PCA plot onto the PC1 axis (i.e. consider the PC1 scores only) and assess whether there is a clear separation between known and unknown APT actors. Then, project onto the PC2 axis (i.e. consider the PC2 scores only) and evaluate whether the separation is better than in PC1. You can access the PCA scores for PC1 and PC2 via mypca$x, assuming mypca contains your PCA results from prcomp(.).

-     the key features in this dimension that can drive this process  (Hint: based on your decision above, examine the loadings from part (v) of your chosen PC and choose those whose absolute loading (i.e. disregard the sign) is greater than 0.3).

Video Presentation Checklist

1. In your video presentation, you must

a. Run your code corresponding to parts (i) to (vii) above

b. Display the relevant output

c. Interpret the output

2. Your video presentation must include a camera shot of yourself in the video

capture, unless there is an exceptional reason and is supported by a Learning  Assessment Plan (LAP). 20% is automatically deducted from your final mark if this is not included in your video presentation. If you choose to record with another application, you must make sure that this feature is included.

3.   Your video presentation must be between 4-5 minutes long.

Marking Rubrics