| Autor: |
Rahimzadeh, Hassan, Sadeghi, Morteza, Mireei, Seyed Ahmad, Ghasemi-Varnamkhasti, Mahdi |
| Zdroj: |
Evolutionary Intelligence; Oct2024, Vol. 17 Issue 5/6, p3291-3301, 11p |
| Abstrakt: |
Due to the non-selective behavior of gas sensors in electronic nose (e-nose) systems, the provided signals in exposure to target analytes contain un-needed information. These are considered as noise reducing the detection accuracy. Feature selection, as a pre-processing step in data analysis, removes extra information from the sensors' signals and provides a more relevant data matrix with lower dimensionality to enhance the system selectivity. In the high-dimensional sensor array response spaces, it is however essential to improve the conventional algorithms to be able to cope with complicated feature selection problem. In this study, in order to acquire optimal responses from the gas sensor array of an e-nose system and increase its selectivity for detection of rice type and its storage duration (freshness), the feature selection problem was formulated in an optimization framework. For this reason, a particle swarm optimization (PSO) with an automatic stagnation detecting system was enhanced by genetic operators of differential evolution. This helped acquiring more exploration ability through providing oriented jumps. It was revealed that the system's detection accuracy was improved when smaller subset of features was utilized instead of the whole response, indicating that the sensor array signals included large amount of irrelevant information. The improved PSO could significantly present lower error values than the standard PSO and other examined conventional algorithms. It was concluded the developed algorithm has the potential to be applied as a promising feature selection algorithm in high-dimensional signals of the e-nose systems. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink