Dimensionality Reduction and feature extraction.pptx

Feature Extraction
Principal Component Analysis (PCA)
Dimensionality Reduction

2
Curse of Dimensionality
 Increasing the number of features will
not always improve classification
accuracy.
 In practice, the inclusion of more
features might actually lead to worse
performance.
 The number of training examples
required increases exponentially with
dimensionality D (i.e., kD). Total: 32 bins
Total: 33 bins
Total: 31 bins
k: number of bins per feature
k=3 bins
per feature

3
3
Dimensionality Reduction
 What is the objective?
Choose an optimum set of
features d* of lower
dimensionality to improve
classification accuracy.
 Different methods can be used
to reduce dimensionality:
Feature extraction
Feature selection
d*

4
4
Dimensionality Reduction (cont’d)
Feature extraction:
computes a new set of
features from the original
features through some
transformation f() .
1
2
1
2
.
.
.
.
.
.
. K
i
i
i
D
x
x x
x
x
x
 
 
   
   
   
   
  
   
   
   
   
 
 
 
x y
1
2
1
2
( )
.
.
.
.
.
.
.
f
K
D
x
x
y
y
y
x
 
 
   
   
   
   
 
 
   
   
   
 
 
 
 
 
x
x y
Feature selection:
chooses a subset of
the original features.
f() could be linear or non-linear
K<<D K<<D

5
1
2
1
2
( )
.
.
.
.
.
.
.
f
K
D
x
x
y
y
y
x
 
 
   
   
   
   
 
 
   
   
   
 
 
 
 
 
x
x y
Feature Extraction
 Linear transformations are particularly attractive because they
are simpler to compute and analytically tractable.
 Given x ϵ RD, find an K x D matrix T such that:
y = Tx ϵ RK where K<<D
5
T
This is a projection
transformation from D
dimensions to K dimensions.
Each new feature yi is a linear
combination of the original
features xi

6
Feature Extraction (cont’d)
 From a mathematical point of view, finding an
optimum mapping y=𝑓(x) can be formulated as an
optimization problem (i.e., minimize or maximize an
objective criterion).
 Commonly used objective criteria:
 Minimize Information Loss: projection in the lower-
dimensional space preserves as much information in the
data as possible.
 Maximize Discriminatory Information: projection in the
lower-dimensional space increases class separability.
6

7
Feature Extraction (cont’d)
 Popular linear feature extraction methods:
 Principal Components Analysis (PCA): Seeks a projection
that minimizes information loss.
 Linear Discriminant Analysis (LDA): Seeks a projection
that maximizes discriminatory information.
 Many other methods:
 Making features as independent as possible (Independent
Component Analysis).
 Retaining interesting directions (Projection Pursuit).
 Embedding to lower dimensional manifolds (Isomap,
Locally Linear Embedding).
7

Feature Extraction
Principal Component
Analysis (PCA)

9
Intuition behind PCA
Find the tallest person.
Person Height
A 185
B 145
C 160
I can by seeing person A is the tallest

10
Intuition behind PCA
Find the tallest person.
Person Height
A 173
B 172
C 174
It’s tough when they are very similar in height.

11
What Is Principal Component Analysis?
 Principal component analysis (PCA) is a
statistical technique to reduce the
dimensionality of complex, high-volume
datasets by extracting the principal
components that contain the most information
and rejecting noise or less important data while
preserving all the crucial details.

12
Principal Component Analysis (PCA)
 PCA comes under the Unsupervised Machine
Learning category
 Reducing the number of variables in a data
collection while retaining as much information
as feasible is the main goal of PCA.
 PCA can be mainly used for Dimensionality
Reduction and also for important feature
selection.
 Correlated features to Independent features

13
 PCA is a method to reduce the dimensionality
of enormous data collections.
 This approach transforms an extensive
collection of variables into a smaller group that
retains nearly all the data in the larger set.
 Lowering the number of variables in a data set
inevitably reduces its precision

14
 The purpose of dimension reduction is to forgo
a certain level of precision for simplification.
 Smaller data collections are easier to
investigate and visualize.
 This facilitates and accelerates the analysis of
data points by machine learning (ML)
algorithms.

15
 PCA is designed to limit the total quantity of
variables in a data set while maintaining as
many details as possible.
 The altered new features or PCA’s results
are known as principal components (PCs)
once PCA has been performed.
 The number of PCs is the same as or lesser
than the number of original features in the
dataset.

16
characteristics of principal components:
1. The principal component must correspond to the
linear arrangement of the initial features.
2. The character of these components is orthogonal.
This indicates there is no relationship within a pair
of variables.
3. From 1 to n, the significance of each component
diminishes.
1. This indicates that the number one PC is the most
important, while the number “n” PC is the least
significant.

17
Basic Terminologies of PCA
 Variance – for calculating the variation of data
distributed across dimensionality of graph
 Covariance – calculating dependencies and
relationship between features
 Standardizing data – Scaling our dataset within a
specific range for unbiased output

18
 Covariance matrix – Used for calculating
interdependencies between the features or
variables and also helps in reduce it to improve
the performance

19
 EigenValues and EigenVectors –
Eigenvectors’ purpose is to find out the largest
variance that exists in the dataset to calculate
Principal Component.
 Eigenvalue means the magnitude of the
Eigenvector.
 Eigenvalue indicates variance in a particular
direction and whereas eigenvector is expanding
or contracting X-Y (2D) graph without altering
the direction.

20
EigenValues and EigenVectors
• In this shear mapping, the blue arrow changes direction whereas the pink
arrow does not.
• The pink arrow in this instance is an eigenvector because of its constant
orientation.
• The length of this arrow is also unaltered, and its eigenvalue is 1.
• Technically, PC is a straight line that captures the maximum variance
(information) of the data.
• PC shows direction and magnitude. PC are perpendicular to each other.

21
 Dimensionality Reduction – Transpose of
original data and multiply it by transposing of
the derived feature vector.
 Reducing the features without losing
information.

22
How does PCA work?
 The steps involved for PCA are as follows-
1. Original Data
2. Normalize the original data (mean =0, variance =1)
3. Calculating covariance matrix
4. Calculating Eigen values, Eigen vectors, and
normalized Eigenvectors
5. Calculating Principal Component (PC)
6. Plot the graph for orthogonality between PCs

24
STEP 1: STANDARDIZATION
 The range of variables is calculated and
standardized in this process to analyze the
contribution of each variable equally.
 Calculating the initial variables will help you
categorize the variables that are dominating the
other variables of small ranges.
 This will help you attain biased results at the
end of the analysis.

25
 To transform the variables of the same
standard, you can follow the following
formula.
 Where,
X= value in a data set N= number of
values in the data set

26
 Example:

27
 Calculate the Mean and Standard Deviation for
each feature and then, tabulate the same as
follows.

28
 after the Standardization of each variable, the
results are tabulated below.

29
STEP 2: COVARIANCE MATRIX COMPUTATION
 In this step, you will get to know how the variables of the
given data are varying with the mean value calculated.
 Any interrelated variables can also be sorted out at the
end of this step.
 To segregate the highly interrelated variables, you
calculate the covariance matrix with the help of the given
formula.
 **Note: **A covariance matrix is a N x N symmetrical matrix
that contains the covariances of all possible data sets.

30
 The covariance matrix of two-dimensional data is, given
as follows:
covariance matrix of 3-dimensional data

31
Insights of covariance matrix
 the covariance of a number with itself is its variance
(COV(X, X)=Var(X)), the values at the top left and
bottom right will have the variances of the same
initial number.
 Covariance Matrix at the main diagonal will be
symmetric concerning the fact that covariance is
commutative (COV(X, Y)=COV(Y, X)).

32
Insights of covariance matrix
 If the value of the Covariance Matrix is positive, :variables are
correlated. ( If X increases, Y also increases and vice versa)
 If the value of the Covariance Matrix is negative: variables are inversely
correlated. ( If X increases, Y also decreases and vice versa).i
 End of this step, we will come to know which pair of variables are
correlated with each other, so that you might categorize them much
easier.

33
 Example: The formula to calculate the covariance
matrix of the given example will be:

34
 Since you have already standardized the features, you can consider
Mean = 0 and Standard Deviation=1 for each feature.
VAR(F1) = ((-1.0695-0)² + (0.5347-0)² + (-1.0695-0)² + (0.5347–0)²
+(1.069–0)²)/5
 On solving the equation, you get, VAR(F1) = 0.78
COV(F1,F2) = ((-1.0695–0)(0.8196-0) + (0.5347–0)(-1.6393-0) + (-
1.0695–0)* (0.0000-0) + (0.5347–0)(0.0000-0)+ (1.0695–0)(0.8196–
0))/5
 On solving the equation, you get, COV(F1,F2 = -0.8586)

35
covariance matrix
 Similarly solving all the features, the
covariance matrix will be,

36
STEP 4: FEATURE VECTOR
 To determine the principal components of
variables, we have to define eigen value and
eigen vectors.
 Let A be any square matrix. A non-zero vector
v is an eigenvector of A if
Av = λv
for some number λ, called the corresponding
eigenvalue.

37
 Once you have computed the eigen vector
components, define eigen values in descending
order ( for all variables) and now you will get a
list of principal components.
 So, the eigen values represent the principal
components and these components represent
the direction of data.

38
 This indicates that if the line contains large
variables of large variances, then there are
many data points on the line. Thus, there is
more information on the line too.
 Finally, these principal components form a line
of new axes for easier evaluation of data and
also the differences between the observations
can also be easily monitored.

39
 Example:
 Let ν be a non-zero vector and λ a scalar.
As per the rule,
 Aν = λν, then λ is called eigenvalue associated
with eigenvector ν of A.

40
 Example:
 Upon substituting the values in det(A- λI) = 0,
you will get the following matrix.

41
 When you solve the following the matrix by considering 0 on
right-hand side, you can define eigen values as
λ = 2.11691 , 0.855413 , 0.481689 , 0.334007
 Then, substitute each eigen value in (A-λI)ν=0 equation and
solve the same for different eigen vectors v1, v2, v3 and v4.
 For instance, For λ = 2.11691, solving the above equation
using Cramer's rule, the values for the v vector are
v1 = 0.515514 v2 = -0.616625 v3 = 0.399314 v4 =0.441098

42
 Follow the same process and you will form the
following matrix by using the eigen vectors
calculated as instructed.

43
 Sort the Eigenvalues in decreasing order.

44
 calculate the sum of each Eigen column, arrange them
in descending order and pick up the topmost Eigen
values.
 These are your Principal components.

45
STEP 5: RECAST THE DATA ALONG THE
PRINCIPAL COMPONENTS AXES
 Still now, apart from standardization, you haven’t made any
changes to the original data.
 You have just selected the Principal components and formed a
feature vector.
 Yet, the initial data remains the same on their original axes.
 This step aims at the reorientation of data from their original
axes to the ones you have calculated from the Principal
components.
 This can be done by the following formula.
Final Data Set= Standardized Original Data Set * FeatureVector

46
Example:
 Standardized
Original Data Set =
 FeatureVector =

47
Example:
 By solving the above equations, you will get
the transformed data as follows.
 Your large dataset is now compressed into a
small dataset without any loss of data!

48
Example in Python
 Step 1: Load the Iris Data Set

49
Step 2: Standardize the Data
• Use StandardScaler to help you standardize the data set’s features
onto unit scale (mean = 0 and variance = 1), which is a requirement
for the optimal performance of many machine learning algorithms.
• If you don’t scale your data, it can have a negative effect on your
algorithm.

50
Step 2: Standardize the Data
• Use StandardScaler to help you standardize the data
set’s features onto unit scale (mean = 0 and variance =
1), which is a requirement for the optimal performance of
many machine learning algorithms.
• If you don’t scale your data, it can have a negative effect
on your algorithm.

51
Step 3: PCA Projection to 2D
 The original data has four columns (sepal length, sepal width,
petal length and petal width).
 In this section, the code projects the original data, which is
four-dimensional, into two dimensions.

52
Step 3: PCA Projection to 2D
 Concatenating DataFrame along axis = 1.
 finalDf is the final DataFrame before plotting the
data.

53
Step 4: Visualize 2D Projection

54
Step 4: Visualize 2D Projection

55
Explained Variance
 The explained variance tells you how much information
(variance) can be attributed to each of the principal
components.
 By using the attribute explained_variance_ratio_, you can
see that the first principal component contains 72.96 percent of
the variance, and the second principal component contains
22.85 percent of the variance.
 Together, the two components contain 95.71 percent of the
information.

56
Perform a Scree Plot of the
Principal Components
 A scree plot is like a bar chart showing the size
of each of the principal components.
 It helps us to visualize the percentage of
variation captured by each of the principal
components.
 To perform a scree plot you need to:
 first of all, create a list of columns
 then, list of PCs
 finally, do the scree plot using plt

Dimensionality Reduction and feature extraction.pptx

More Related Content

What's hot (20)

Similar to Dimensionality Reduction and feature extraction.pptx (20)

More from Sivam Chinna (10)

Recently uploaded (20)

Dimensionality Reduction and feature extraction.pptx