rbs - presentation about applications of machine learning.

Applications of Machine Learning
for Interdisciplinary Research
Dr. R. Balasundaram, M.E., Ph.D.,
Professor
Department of Mechanical Engineering,
SRM Institute of Science and Technology ,
Tiruchirappalli Campus – 621105
E-mail : balasundaram_rbs@yahoo.co.in 1

Topics to be covered
What is Data Science ?- some examples
DT Algorithm introduction and application
Numerical illustration of Decision Tree algorithm for
manufacturing dataset
Association Rule analysis
Other Areas of implementation
2

Metrological Applications
6
Direction of Michaung - Cyclone

Algorithms
7
Source : Machine learning approaches for prediction of properties of
natural fiber composites : Apriori Algorithm
Australian Journal of Mechanical Engineering
https://www.tandfonline.com/doi/full/10.1080/14484846.2022.2030091

Example for Decision Tree
Sl. No. Gender Age Blood Pressure Drug
1 Male 20 Normal A
2 Female 43 Normal B
3 Male 37 High A
4 Male 33 Low B
5 Female 48 High A
6 Male 29 Normal A
7 Female 52 Normal B
8 Male 42 Low B
9 Male 61 Normal B
10 Female 30 Normal A
11 Female 26 Low B
12 Male 54 High A
Input variables : (predictor attributes) – Gender, Age, BP
Output Variables : (class attribute / target variable ) : Type of Drug
Source : Data mining : Theory and Practice by K.P. Soman
8

 Highest value of entropy will act as “ root node “
 The rules from the tree are
i) If BP is High then prescribe Drug ‘A’ (accuracy – 3/3 ,100% Sl.No. 3, 5,12))
ii)If BP is low then Prescribe Drug ‘B’ etc.,( accu – 3/3, 100% 4,8,11)
III) if BP is normal and AGE <= 40 , Drug ‘A” (accu- 3/3 100% , 1,6,10)
Iv) if BP is normal and AGE >=40 Drug ‘B’ ( accur – 3/3, 100% 2,4,9)
The Overall Accuracy of the tree is 100 %
9

Decision Tree Prediction of wear rate for zinc
oxide filled AA7075 matrix composites
10
Taguchi and Decision Tree Approach Algorithm for Prediction of wear rate in ZnO filled
AA7075 Matrix Composite
Published : Surface Topography – Metrology and Properties , vol.9, 2021

Instance No.
Input variables (predictor attributes) Output variable (class attribute)
Reinforcement (wt.%) Applied Load(N)
Sliding Velocity
(m/s)
Sliding Distance (m) Wear Rate (mm3/m)
1 0 10 1 1000 0.00294
2 0 10 2 1500 0.00343
3 0 10 3 2000 0.00386
4 0 20 1 1500 0.00318
5 0 20 2 2000 0.00367
6 0 20 3 1000 0.00441
7 0 30 1 2000 0.00514
8 0 30 2 1000 0.00477
9 0 30 3 1500 0.00465
10 5 10 1 1500 0.00266
11 5 10 2 2000 0.00309
12 5 10 3 1000 0.00254
13 5 20 1 2000 0.00327
14 5 20 2 1000 0.00315
15 5 20 3 1500 0.00290
16 5 30 1 1000 0.00347
17 5 30 2 1500 0.00363
18 5 30 3 2000 0.00381
19 10 10 1 2000 0.00179
20 10 10 2 1000 0.00143
21 10 10 3 1500 0.00167
22 10 20 1 1000 0.00215
23 10 20 2 1500 0.00239
24 10 20 3 2000 0.00269
25 10 30 1 1500 0.00335
26 10 30 2 2000 0.00359
27 10 30 3 1000 0.00287
Wear rate of 27 Samples
12

DT dataset :
Output must be categorical one
Yes /No
A, B, C ..
Low , Medium , High
Input parameters :
Either Categorical or Numerical
13

Significance of DT
• It converts the low level data in to High level knowledge
• The Rules are easy to understand
• Need not depends experts
• It is type of classification algorithm to classify the given data
set
(For Decision Tree – ID3, C4.5, CART)
C4.5 is most powerful
14

Association Rule (Market Basket)
Analysis
15

Dataset
16
It will not accept numerical input

Case study –
Quality tests of 15 engines
Authors used “Rapid Miner “ software
And paper is published in information science and applications
17

Case study- process control
. The objectives of the research are to design and verify the data mining tools
to support production process control for decision making,. The authors used
CRISP – DM tool
18

Data Mining concepts can be
implemented in
• Group Technology – clustering of components
• Analysis of engine performance
• Analysis of boiler – stream of data (similar to net
browsing)
• Analysis of sensors outputs – stream of data
• Analysis of solar panel performance
• Combination of two or more algorithms ( like DM +
Non traditional optimization)
19

Algorithms
 Classification (DT)
 Regression ( NN, SVM)
 Association (Apriori, FP Growth)
 Clustering ( KNN)
Current trend:
 Hybrid of DM + Optimization
 Wavelets
 Image Processing
 Fractals
20

Thank you
21
Contact : balasundaram_rbs@yahoo.co.in

rbs - presentation about applications of machine learning.

More Related Content

Similar to rbs - presentation about applications of machine learning. (20)

Recently uploaded (20)

rbs - presentation about applications of machine learning.