0% found this document useful (0 votes)
18 views26 pages

ML Model Evaluation Techniques Guide

Evaluation of Classification

Uploaded by

jusinbee
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
18 views26 pages

ML Model Evaluation Techniques Guide

Evaluation of Classification

Uploaded by

jusinbee
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Data Mining:

Concepts and Techniques

Chapter 7 —ML Model Evaluation Techniques

Dr. Yatendra Sahu


Assistant Professor, Department of CSE
Indian Institute of Information Technology Bhopal

1
Training To build
Dataset model

Dataset
To
Test Evaluate
Dataset Model
Training set:
 Training set is used to build a model.
 It is used find relevant information on how to associate input data with
output decision. The system is trained by applying these algorithms on the
dataset, all the relevant information is extracted from the data and results are
obtained.
 Generally, 70% of the data of the dataset is taken for training data.
Testing set:
 Testing data is used to test model. It is the set of data which is used to verify
whether the system is producing the correct output after being trained or
not. Generally, 30% of the data of the dataset is used for testing.
2
Learning Algorithm
 Machine Learning is a concept which provides ability to the machine to
automatically learn and improve from experience without being explicitly
programmed.
 The process of learning begins with observations in order to find patterns in
data and make better decisions in the future based on the examples that we
provide.
 The primary aim of learning algorithm is to allow the computers learn
automatically without human intervention

Machine Learning
Algorithm

Supervised Un-Supervised Reinforcement


Learning Learning Learning
Algorithm Algorithm Algorithm 3
Types of Supervised Learning
Machine Learning
Algorithm

Supervised Un-Supervised Reinforcement


Learning Learning Learning
Algorithm Algorithm Algorithm

Regression Classification

4
10
Num-1 Num-2 Sum

5 5 10
8 2 10 5
Model Logic
10 3 13
5
15 6 21
20 4 21 Training Phase
30 40 70

30
Trained Model 70
40

Testing Phase

5
Evaluation Metric for Regression and
Classification

6
Binary Classification Multi-Class Classification

Source::
[Link]
DISEASE_PREDICTION_SYSTEM_HDPS/figures?lo=1

7
Classification Evaluation Metric
• True Positives (TP): Actual TRUE, which was predicted as TRUE
• True Negatives (TN): Actual FALSE, which was predicted as FALSE

• False Positives (FP): Actual FALSE, which was predicted as TRUE (Type I
error)
• False Negatives (FN): Actual TRUE, which was predicted as FALSE (Type
II error)

1) Confusion Matrix

 A confusion matrix is created by comparing the predicted class label of a


data point with its actual class label.

 A confusion matrix can be created for a binary classification as well as a


multi-class classification model. 8
Actual Predicted
Output (y) Output (y’) 1 0
1 3 1
1 1
1 1 1 1
0
0 1
0 0
1 0
1 1

9
 It is a n*n matrix, where n is a number of classes.

10
[Link]
11
12
Why Confusion Matrix?

13
Accuracy
 It is the proportion of the total number of predictions that are correct.

Accuracy
= (45+30)/(45+20+5+30) = 75%

The 75% of examples are correctly


classified by the classifier.

Normal Fraud

Normal
940 10

40 10
Fraud
14
Accuracy/Precision/ Recall

Source: [Link]
15
Precision
 It is a ratio of total number of correctly classified positive examples and the
total number of predicted positive examples. It shows correctness achieved
in positive prediction.

 Precision becomes important in cases where we are more concerned about


finding the maximum number of positive class even if the total accuracy
reduces.

Precision = TP/(TP+FP)

Precision = 45/(45+5)= 90%

The 90% of examples are classified as spam


are actually spam.

16
Recall / Sensitivity / True Positive Rate

 It is measure of positive examples labelled as positive by classifier. It


should be higher.

Recall= TP/(TP+FN) = TP/P

Sensitivity = 45/(45+20) = 69.23% .

The 69.23% spam emails are correctly


Classify

17
Specificity / True Negative Rate

 It is measure of negative examples labeled as negative by classifier.


There should be high specificity.

Specificity = TN/(TN+FP) = TN/N

specificity = 30/(30+5) = 85.71%

The 85.71% non-spam emails


are accurately classified

18
F1 Score

 It is a weighted average of the recall and precision.

 F1 score might be good choice when you seek to balance between


Precision and Recall.

19
20
How to Calculate Accuracy, Precision
and Recall for More than Two Classes

21
Source: [Link] 22
Source: [Link] 23
24
Source: [Link] 25
Education – Partnership – Solutions

26

You might also like