Genetic algorithms (GAs) and evolutionary strategy to optimize electronic nose sensor selection

C. Li, Paul Heinz Heinemann, P. M. Reed

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

The high dimensionality of electronic nose data increases the difficulty of their use in classification models. Reducing this high dimensionality helps reduce variable numbers, potentially improve classification accuracy by removing irrelevant sensors, and reduce computation time and sensor cost. In this research, the Cyranose 320 electronic nose was optimized for apple defect detection by selecting the most relevant of its 32 internal sensors using various selection methods. The contribution of each sensor was first evaluated statistically by calculating the F-value. By keeping only the top 90% cumulative F-values, 25 sensors were selected and the classification error rate was 25.4%. Sequential forward/backward search methods reduced the minimum classification error rate to 6.1%. Two more heuristic optimization algorithms, genetic algorithm (GA) and the covariance matrix adaptation evolutionary strategy (CMAES), were applied and compared. Although both algorithms gave a best classification error rate of 4.4%, the average classification error rate of CMAES over 30 random seed runs was 5.0% (SD = 0.006), which was better than the 5.2% (SD = 0.004) from the GA. The final optimal solution sets obtained by using an integer GA showed that including more sensors did not guarantee better classification performance. The best reduction in classification error rate was 10%, while the number of sensors was reduced by 75%. This study provided a robust and efficient optimization approach to reduce the high dimensionality of electronic nose data, which substantially improved electronic nose performance in apple defect detection while potentially reducing the overall electronic nose cost for future specific applications.

Original languageEnglish (US)
Pages (from-to)321-330
Number of pages10
JournalTransactions of the ASABE
Volume51
Issue number1
StatePublished - Jan 1 2008

Fingerprint

Electronic Nose
electronic nose
genetic algorithm
sensors (equipment)
Genetic algorithms
sensor
Sensors
evolutionary adaptation
Malus
Covariance matrix
defect
apples
Costs and Cost Analysis
matrix
electronics
Electronic nose
selection methods
heuristics
cost
Seed

All Science Journal Classification (ASJC) codes

  • Forestry
  • Food Science
  • Biomedical Engineering
  • Agronomy and Crop Science
  • Soil Science

Cite this

@article{d0277d57b4b3425db6df50258412cab2,
title = "Genetic algorithms (GAs) and evolutionary strategy to optimize electronic nose sensor selection",
abstract = "The high dimensionality of electronic nose data increases the difficulty of their use in classification models. Reducing this high dimensionality helps reduce variable numbers, potentially improve classification accuracy by removing irrelevant sensors, and reduce computation time and sensor cost. In this research, the Cyranose 320 electronic nose was optimized for apple defect detection by selecting the most relevant of its 32 internal sensors using various selection methods. The contribution of each sensor was first evaluated statistically by calculating the F-value. By keeping only the top 90{\%} cumulative F-values, 25 sensors were selected and the classification error rate was 25.4{\%}. Sequential forward/backward search methods reduced the minimum classification error rate to 6.1{\%}. Two more heuristic optimization algorithms, genetic algorithm (GA) and the covariance matrix adaptation evolutionary strategy (CMAES), were applied and compared. Although both algorithms gave a best classification error rate of 4.4{\%}, the average classification error rate of CMAES over 30 random seed runs was 5.0{\%} (SD = 0.006), which was better than the 5.2{\%} (SD = 0.004) from the GA. The final optimal solution sets obtained by using an integer GA showed that including more sensors did not guarantee better classification performance. The best reduction in classification error rate was 10{\%}, while the number of sensors was reduced by 75{\%}. This study provided a robust and efficient optimization approach to reduce the high dimensionality of electronic nose data, which substantially improved electronic nose performance in apple defect detection while potentially reducing the overall electronic nose cost for future specific applications.",
author = "C. Li and Heinemann, {Paul Heinz} and Reed, {P. M.}",
year = "2008",
month = "1",
day = "1",
language = "English (US)",
volume = "51",
pages = "321--330",
journal = "Transactions of the ASABE",
issn = "2151-0032",
publisher = "American Society of Agricultural and Biological Engineers",
number = "1",

}

Genetic algorithms (GAs) and evolutionary strategy to optimize electronic nose sensor selection. / Li, C.; Heinemann, Paul Heinz; Reed, P. M.

In: Transactions of the ASABE, Vol. 51, No. 1, 01.01.2008, p. 321-330.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Genetic algorithms (GAs) and evolutionary strategy to optimize electronic nose sensor selection

AU - Li, C.

AU - Heinemann, Paul Heinz

AU - Reed, P. M.

PY - 2008/1/1

Y1 - 2008/1/1

N2 - The high dimensionality of electronic nose data increases the difficulty of their use in classification models. Reducing this high dimensionality helps reduce variable numbers, potentially improve classification accuracy by removing irrelevant sensors, and reduce computation time and sensor cost. In this research, the Cyranose 320 electronic nose was optimized for apple defect detection by selecting the most relevant of its 32 internal sensors using various selection methods. The contribution of each sensor was first evaluated statistically by calculating the F-value. By keeping only the top 90% cumulative F-values, 25 sensors were selected and the classification error rate was 25.4%. Sequential forward/backward search methods reduced the minimum classification error rate to 6.1%. Two more heuristic optimization algorithms, genetic algorithm (GA) and the covariance matrix adaptation evolutionary strategy (CMAES), were applied and compared. Although both algorithms gave a best classification error rate of 4.4%, the average classification error rate of CMAES over 30 random seed runs was 5.0% (SD = 0.006), which was better than the 5.2% (SD = 0.004) from the GA. The final optimal solution sets obtained by using an integer GA showed that including more sensors did not guarantee better classification performance. The best reduction in classification error rate was 10%, while the number of sensors was reduced by 75%. This study provided a robust and efficient optimization approach to reduce the high dimensionality of electronic nose data, which substantially improved electronic nose performance in apple defect detection while potentially reducing the overall electronic nose cost for future specific applications.

AB - The high dimensionality of electronic nose data increases the difficulty of their use in classification models. Reducing this high dimensionality helps reduce variable numbers, potentially improve classification accuracy by removing irrelevant sensors, and reduce computation time and sensor cost. In this research, the Cyranose 320 electronic nose was optimized for apple defect detection by selecting the most relevant of its 32 internal sensors using various selection methods. The contribution of each sensor was first evaluated statistically by calculating the F-value. By keeping only the top 90% cumulative F-values, 25 sensors were selected and the classification error rate was 25.4%. Sequential forward/backward search methods reduced the minimum classification error rate to 6.1%. Two more heuristic optimization algorithms, genetic algorithm (GA) and the covariance matrix adaptation evolutionary strategy (CMAES), were applied and compared. Although both algorithms gave a best classification error rate of 4.4%, the average classification error rate of CMAES over 30 random seed runs was 5.0% (SD = 0.006), which was better than the 5.2% (SD = 0.004) from the GA. The final optimal solution sets obtained by using an integer GA showed that including more sensors did not guarantee better classification performance. The best reduction in classification error rate was 10%, while the number of sensors was reduced by 75%. This study provided a robust and efficient optimization approach to reduce the high dimensionality of electronic nose data, which substantially improved electronic nose performance in apple defect detection while potentially reducing the overall electronic nose cost for future specific applications.

UR - http://www.scopus.com/inward/record.url?scp=41749101059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=41749101059&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:41749101059

VL - 51

SP - 321

EP - 330

JO - Transactions of the ASABE

JF - Transactions of the ASABE

SN - 2151-0032

IS - 1

ER -