Preprocess

Question 1

The 6 steps of CRISP-DM

Answer 1

1. Business Understanding
2. Data understanding
3. Data preparation
4. Modelling
5. Evaluation
6. Deployment

Question 2

EDA GOAL

Answer 2

Frame hypothesis, visualies, discover patterns, spot anomalies

Question 3

3 steps of EDA

Answer 3

1. Generate questions
2. Search for the answers with visualising, transforming, modelling.
3. Refine/generate new questions with the answers

Question 4

One-hot enconding

Answer 4

For each different value, a new feature is added and the values is either 1 or 0 for the feature.

Question 5

Normalisation

Answer 5

KNN, SVM, gradient descent are affected by scale of data. Normalisation brings to center around 0. Kinda makes it normally distributed.

Question 6

Standardisation

Answer 6

Use if the distribution is normal. Brings the data to the same range.

Question 7

Discretisation

Answer 7

Bring continuous values to discrete. E.g. for decision tree you would need discrete values.

Question 8

MAR

Answer 8

Missing data depends on other observed variables but not on the value of the missing data. e.g. missing income depends on the education level but not on the income itself.

Question 9

MNAR

Answer 9

e.g. high income individuals are less likely to disclose their income. The missing data depends on the missing values.

Question 10

DELETION

Answer 10

• When less than 5% contains missing values
• Or MCAR

Question 11

The curse of dimensionality:

Answer 11

High dimensional data has irrelevant and redundant features. High correlation between features.

Question 12

Filter FS

Answer 12

uses existing measures, no learning algorithm

Question 13

Wrapper FS

Answer 13

Uses learning performance to train ML models many times

Question 14

Embedded FS

Answer 14

Train the ML model once. Select features based on learned model.

Question 15

Feature Raking

Answer 15

Evaluate features individually
rank features and select top-ranked feautres
Simple, efficient
But ignores feature interactions (it can select redundant features)

Question 16

Feature Subset Selection

Answer 16

Iteratively evaluate the whole feature subset. Considers feature interactions and ahs better performance but is more expensive/complicated

Question 17

SFS steps

Answer 17

1. Start with emptey feature set
2. Sequentially add the feature X that results in the highest value when combined with the current set.
3. Stop when the defined N amount of features is selected.
   (works best when the optimal subset has a small N number of features)

Question 18

SBFS steps

Answer 18

1. start from the full set
2. sequentially remove the feature X that results in the highets objective value
3. Stop when a pre-defined number of features is selected.
   (Works best when the N number of features is large)

Question 19

Filter FC

Answer 19

Class. Accuracy: low, computational cost: low, generality: high

Question 20

Embedded FC

Answer 20

Class. acc: medium, comp. cost: medium, Generality : medium

Question 21

Wrapper FC

Answer 21

Class. acc: high, Comp. cost: high, Generality: low.

Question 22

PCA

Answer 22

a mathematical procedure that linearly transforms
(possibly) correlated features into a (smaller) number of
uncorrelated features called principal components.
Goal is to achieve high data variance

Question 23

What is data variance? Why higher variance?

Answer 23

Variance: squared deviations from the mean. Spread of your dataset. We want higher varaince because we want to retain as much informationa s possible.