The Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things Environments

1. Introduction Background Evaluation Methodology Results Conclusion
The Elimination-Selection Based Algorithm for
Efficient Resource Discovery in Internet of Things
Environments
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and
Stephan Reiff-Marganiec
January 2018
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 1 / 36

2. Introduction Background Evaluation Methodology Results Conclusion
Introduction
What do you think when we talk about Internet of Things?
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 2 / 36

3. Introduction Background Evaluation Methodology Results Conclusion
Introduction
What do you think when we talk about Internet of Things?
Figure: Smart home
Figure: Smart city
Figure: Smart car
Figure: Smart farm
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 3 / 36

4. Introduction Background Evaluation Methodology Results Conclusion
Introduction
Cloud of Things
Development of unprecedented services
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 4 / 36

5. Introduction Background Evaluation Methodology Results Conclusion
Introduction
Cloud of Things
Development of unprecedented services
Models
Data as a Service
Sensor as a Service
Sensing as a Service
Big data as a Service
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 5 / 36

6. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Figure: Solutions to provide sensor and sensor data as a service
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 6 / 36

7. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Let’s take the Flamingo Las Vegas as example
Figure: Restaurants near to Flamingo Las Vegas
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 7 / 36

8. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Let’s take the Flamingo Las Vegas as example
Figure: Restaurants near to Flamingo Las Vegas
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 8 / 36

9. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Let’s take the Flamingo Las Vegas as example
Figure: Restaurants near to Flamingo Las Vegas
Static criteria
Distance
Price
Dynamic criteria
Service time in the last hour
Quality of service in the past four
hours
Percentage of available tables in the
last 10 minutes
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 9 / 36

10. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Problem
A resource(restaurant) could not be suitable to accomplish the requirements of other
users
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 10 / 36

11. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Problem
A resource(restaurant) could not be suitable to accomplish the requirements of other
users
Resource discovery
Context
User constraints
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 11 / 36

12. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Problem
A resource(restaurant) could not be suitable to accomplish the requirements of other
users
Resource discovery
Context
User constraints
Solution
Multiple Criteria Decision Analysis
Context Aware Computing
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 12 / 36

13. Introduction Background Evaluation Methodology Results Conclusion
Related Works
Figure: Related works focused on techniques
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 13 / 36

14. Introduction Background Evaluation Methodology Results Conclusion
Related Works
Gap
These solutions are concerned just with the selection time of the algorithms. They not
consider the quality of the proposed solution or the impact of user constraints in their
quality.
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 14 / 36

15. Introduction Background Evaluation Methodology Results Conclusion
Motivation
Objective
We propose a novel Elimination-Selection (E-S) algorithm to increase the quality of the
proposed resources in a timmely manner.
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 15 / 36

16. Introduction Background Evaluation Methodology Results Conclusion
Multiple-criteria decision-analysis algorithms
Multiples Objective Optimization
Different comparable units
Assist the judgment of decision making
Goals and criteria
The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)
Vector normalisation
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 16 / 36

17. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
TOPSIS
Normalize the available options
Distance (min)
Price(min)
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 17 / 36

18. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
TOPSIS
Compute the Ideal Positive Solution and Ideal Negative Solution
Distance (min)
Price(min)
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 18 / 36

19. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
TOPSIS
The solution which is closer to Ideal Positive Solution and farthest from Negative
Solution is the best one
Distance (min)
Price(min)
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 19 / 36

20. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
Pareto Optimality
Compare different sets of solutions through
dominance.
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 20 / 36

21. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
Pareto Optimality
Compare different sets of solutions through
dominance.
Condition 1
The x solution could not be worse than y for
all objectives
f(x) ≥ f(y), for all objectives
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 21 / 36

22. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
Pareto Optimality
Compare different sets of solutions through
dominance.
Condition 1
The x solution could not be worse than y for
all objectives
f(x) ≥ f(y), for all objectives
Condition 2
The x solution must be better than y at list
for one objective
f(x) > f(y), at least for one objective
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 22 / 36

23. Introduction Background Evaluation Methodology Results Conclusion
Multi-Objective optimization
Pareto Optimality
Compare different sets of solutions through
dominance.
Condition 1
The x solution could not be worse than y for
all objectives
f(x) ≥ f(y), for all objectives
Condition 2
The x solution must be better than y at list
for one objective
f(x) > f(y), at least for one objective
Fast Non Dominated Sort [2]
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 23 / 36

24. Introduction Background Evaluation Methodology Results Conclusion
Elimination Sort Algorithm
Table: List of variable used in the algorithm
Variable Description
Q Available options
N Number of options to be selected
SR Search rate value
S Selected options
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 24 / 36

25. Introduction Background Evaluation Methodology Results Conclusion
Evaluation Workflow
Table: Approximately weather conditions required to seed a corn crop based on Doblas-Reyes et al. [3] and Rosen-
zweig et al. [5] al.
Factor Value
Temperature (t) 59 ºF <t <91.4 ºF
Humidity (h) h >50 %
Sky coverage (sc) sc >50 %
Wind (w) w <20 km/h
Precipitation (p) 0 mm <p <50 mm
Table: Objectives functions to seed a corn crop
Context property Objective Function
Temperature max(−0.01777778t2 + 10.512t − 1552.9364)
Humidity max(h)
Sky coverage max(sc)
Wind min(w)
Precipitation max(0.0016p2 + 0.08p)
Datetime (dt) max(dt)
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 25 / 36

26. Introduction Background Evaluation Methodology Results Conclusion
Evaluation Workflow
Figure: Evaluation Workflow. Adapted from Nunes et al. [4]
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 26 / 36

27. Introduction Background Evaluation Methodology Results Conclusion
Factors and levels
Table: Factors and levels used in the experiment
Factor Level
Number of Resource Descriptions 209,555
MCDA Method TOPSIS and E-S
Number of Selected Sensors 2095
Number of User Constraint 2, 4 and 6
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 27 / 36

28. Introduction Background Evaluation Methodology Results Conclusion
Metric of Evaluation
Overall nondominated vector generation metric
Shows the number of the proposed MCDA solutions as a proportion of the number of
solutions in the Pareto front [1]
ONVGRfront(n) =
MCDAfront(n)
Paretofront(n)
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 28 / 36

29. Introduction Background Evaluation Methodology Results Conclusion
First Phase
2 Criteria 4 Criteria 6 Criteria
01000020000300004000050000
#ofResources
# of Criteria
46755
14489 14394
Figure: Number of resources considering the conditions to seed a corn crop
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 29 / 36

30. Introduction Background Evaluation Methodology Results Conclusion
Quality
q q q q q q
q
q
q q
q
q q q
q
q
q
q q
q
q
q
q
q
q
q
q
0 10 20 30 40 50 60 70
0.00.20.40.60.81.0
ONVGR
# Fronts
Algorithms
Topsis E−S Optimal
Figure: ONVGR value for 2 criteria
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 30 / 36

31. Introduction Background Evaluation Methodology Results Conclusion
Quality
q
q
q
q
q
2 4 6 8
0.00.20.40.60.81.0
ONVGR
# Fronts
Algorithms
Topsis E−S Optimal
Figure: ONVGR value for 4 criteria
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 31 / 36

32. Introduction Background Evaluation Methodology Results Conclusion
Quality
q q
1 2 3 4 5
0.00.20.40.60.81.0
ONVGR
# Fronts
Algorithms
Topsis E−S Optimal
Figure: ONVGR value for 6 criteria
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 32 / 36

33. Introduction Background Evaluation Methodology Results Conclusion
Time
TOPSIS E−S Optimal TOPSIS E−S Optimal TOPSIS E−S Optimal
0200400600
Time(s)
2 Criteria 4 Criteria 6 Criteria
Algorithm
# of Criteria
Figure: Time to select the resources regarding the conditions
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 33 / 36

34. Introduction Background Evaluation Methodology Results Conclusion
Conclusion
General Conclusion
The number of criteria impacts directly on the number of available resources in the
query phase and the ONVGR value
E-S algorithm conclusion
E-S algorithm show the ONVGR metric value at least equal than TOPSIS
algorithm
E-S algorithm present an acceptable tradeoff between quality and time
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 34 / 36

35. Introduction Background Evaluation Methodology Results Conclusion
End
Thank you!
Questions?
lhnunes@icmc.usp.br
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 35 / 36

36. Introduction Background Evaluation Methodology Results Conclusion
References I
Y. Collette and P. Siarry.
Multiobjective Optimization.
Springer Berlin Heidelberg, 2004.
K. Deb.
Multi-objective optimization using evolutionary algorithms.
John Wiley & Sons, Chichester New York, 2001.
F. Doblas-Reyes, A. Garcia, J. Hansen, L. Mariani, A. Nain, K. Ramesh,
L. Rathore, and R. Venkataraman.
Weather and climate forecasts for agriculture.
Guide to agricultural, meteorological practices, page 57, 2003.
L. H. Nunes, J. C. Estrella, C. Perera, S. Reiff-Marganiec, and A. N. Delbem.
Multi-criteria iot resource discovery: A comparative analysis.
Software: Practice and Experience, -(-):–, 2016.
In print.
C. Rosenzweig, A. Iglesias, X. Yang, P. R. Epstein, and E. Chivian.
Climate change and extreme weather events; implications for food production,
plant diseases, and pests.
Global change & human health, 2(2):90–104, 2001.
Luiz H. Nunes, Júlio C. Estrella, Alexandre C. B. Delbem, Charith Perera and Stephan Reiff-MarganiecThe Elimination-Selection Based Algorithm for Efficient Resource Discovery in Internet of Things EnvironmentsJanuary 2018 36 / 36