Question $8/$ Vraag $8$

In order to apply machine learning to a particular prob$-$

lem, your team collected $12000$ data points of $10$ input

features from We$-$Buy$-$Cars, which is a large online car

sales platform. The car prices collected serve as target

fearure.

$($a$)$ One of the most commonly used processes for pre$-$

dictive data analytics projects is the Cross Industry

Standard Process for Data Mining $($CRISP$-$DM$).$

Name and describe the first $3$ of the $6$ key phases of

the predictive data analytics project life$-$cycle that

are defined by the CRISP$-$DM process.

$($b$)$ A predictive model is to be developed, which

anticipates the price of a car. Is this an example

of supervised or unsupervised learning? Motivate

your answer.

$[$c$)$ Two of the features in the database

are fue$1\_$tank$\_$capacity and the

fue$1\_$consumption, which are both strongly

correlated with expected fuel consumption.

To reduce data set cardinality, your team de$-$

cides to replace fue$1\_$tank$\_$capacity

as a categorical feature, which consists of

$4$ values: $[$smal$1,$ medium, large,

extra$-$large$].$ To categorise each data en$-$

try for fue$1\_$tank$\_$espacity, would you use

$K-$Means clustering or $KNN$ classification? Moti$-$

vate your answer.

$($d$)$ If you want to analyse specific properties of a data

subset, which has an unknown ordering, should

you use top sampling or random sampling? Give a

reason for your answer.