Lecture 7 - Features

Question 1

What are the 4 stages of data pre-processing

Answer 1

1. Data cleaning
2. Data integration
3. Data reduction
4. Data transformation

Question 2

What are features

Answer 2

features, also called attributes, are defined as mapping from the instance space to the feature domain.

Question 3

What are the three main categories of feature statistics

Answer 3

• Statistics of central tendency
• Statistic of dispersion
• Shape statistics

Question 4

What are the 3 main statistics of central tendency

Answer 4

1. mean
2. median
3. mode

Question 5

What are the 2 statistics of dispersion

Answer 5

1. Variance omega^2
2. Standart deviation omega

Question 6

What are the statistics of dispersion

Answer 6

• range
• midrange point
• quantiles
• interguartile range

Question 7

The ____ is more sensetive to outlier than the ____

median or mean

Answer 7

mean
median

Question 8

what is skewness

Answer 8

Skewness is then defined as m/omega^3. A positive value of skewness means that the distribution is right-skewed, which means that the right tail is longer than the left tail. Negative skweness indicates the opposite.

Question 9

What is Kurtosis

Answer 9

m/omega^4. People often use excess kurtosis m/omega^4 - 3. Positive excess kurtosis means that the distrubution is more sharply peaked than the normal distribution.

Question 10

when can structured features be constructed

Answer 10

1. prior to learning the model
2. during learning the model

Question 11

What is normalisation

Answer 11

From Quantitave to Quantitative
Adapt the scale of quantitative features.

Question 12

What is calibration

Answer 12

From ortinal, categorical and boolean TO Quantitative
Adds a scale to features that don’t have one

Question 13

What is discretisation

Answer 13

• from quantitative to ordinal
• from quantitative to categorical

Question 14

what is ordering

Answer 14

• from ordinal to ordinal
• from categorical to ordinal
• from boolean to ordinal

Question 15

What is unordering

Answer 15

from ordinal to categorical

Question 16

what is grouping

Answer 16

from categorical to categorical

Question 17

what is thresholding

Answer 17

• from quantitative to boolean
• from ordinal to boolean

Question 18

what is binarisation

Answer 18

from categorical to boolean

Question 19

Define thresholding. in words not table

Answer 19

Thresholding transforms a quantitave or an ordinal feature into a boolean feature by dinding a feature value to split on.

Question 20

how do we set the threshold for thresholding?

Answer 20

• Supervised thresholding: hand picked for better performance
• unsupervised thresholding: use centeral tendency statistics like mean/median

Question 21

Describe Discretisation

Answer 21

Discretisation transforms a quantitative feature into an ordinal feature, by creating bins where each bin is an interval

Question 22

name and exaplain 2 types of discretisation

Answer 22

• supervised: bottom-up, work by progressively splitting bins
• unsupervised: equal bin width, equal width discretisation

Question 23

Define normalisation

Answer 23

Feature normalisation neutralises the effect of different quantitative features being measured on different scales.

Question 24

Give to formulas with which we can normalise data

Answer 24

• min-max
• z-scores

Question 25

what is PCA

Answer 25

Principal component analysis is a feature-construbtion teqnique. It works by computing the principal components and using them to performs a change of basis on the data.

Question 26

Can PCA be performed on quantitative features?

Answer 26

yes

Question 27

What is the idea of pca

Answer 27

The idea of PCA is to find tehse correlcations and create a new feature that could be represented as a linear combination of the oringial features.

Question 28

in PCA, the sum of squared distances of projected points from the origin are called ____

Answer 28

eigenvalues

Question 29

What are principal components

Answer 29

principal components are new features constructed as a linear combination of original features

Question 30

give 2 approaches to extract principal components

Answer 30

1. Singular value decomposition 2. eigendecomposition

Question 31

How does singular value decomposition work

Answer 31

using matrixs rows for each feature.

Question 32

What is imputation

Answer 32

Imputation is the process of filling in missing data

Question 33

name 3 imputation techniques

Answer 33

1. Mean imputation 2. Regression imputation 3. Expectation maximisation

Question 34

what is mean imputation

Answer 34

calculate the per class mean/median/mode

Question 35

what is regression imputation

Answer 35

a regression model is estimated to predict the observed vlaues of a variable based on other variables.

Question 36

what is expectation maximisation

Answer 36

assuming a multivariate model over all features, use the observed values for maximum-likelyhood estimation of the model parameters, then derive expectations for the unobserved feature values and interate.