Identification of Complex Samples Based on Broad Learning System and Physicochemical Indicators
10.19756/j.issn.0253-3820.251080
- VernacularTitle:基于宽度学习系统和理化指标的复杂样品鉴别方法研究
- Author:
Jia-Qi XIE
1
;
Qiang ZHANG
;
Pei-Ran LIU
;
Ya-Fei YANG
;
Xi-Hui BIAN
Author Information
1. 天津工业大学,天津市绿色化工过程工程重点实验室,天津 300387
- Keywords:
Broad learning system;
Physicochemical indicators;
Complex sample;
Identification;
Confusion matrix
- From:
Chinese Journal of Analytical Chemistry
2025;53(6):944-954,中插16-中插21
- CountryChina
- Language:Chinese
-
Abstract:
Compared to traditional machine learning algorithms,which often suffer from low feature extraction efficiency,insufficient nonlinear pattern recognition capabilities and slow training speeds,the broad learning system(BLS)enhances the learning ability and efficiency by horizontally expanding the network structure.BLS offers advantages such as a simple structure,fast training speed,and strong generalization capabilities.While BLS has demonstrated potential in various fields,but its application in identification of complex samples has not been fully explored.This research investigated the feasibility of using BLS algorithm for identification of complex samples based on physicochemical indicators.Using the iris,wine,and breast cancer datasets,the length and width of petals and sepals of iris flowers,the physicochemical properties of wine,and the nuclear characteristics of breast cancer cells were used as input variables to establish BLS models for identifying iris species,wine varieties,and benign versus malignant nucleus.The model performance was evaluated by confusion matrices,accuracy,and runtime.Compared with partial least squares-discriminant analysis(PLS-DA),soft independent modeling of class analogies(SIMCA),and artificial neural networks(ANN),the results indicated that BLS demonstrated significant advantages in computational efficiency and recognition accuracy.Thus,BLS was an efficient and reliable method for identification of complex samples.
