近红外光谱分析中的变量选择算法研究进展

发布时间：2018-10-25 10:29

【摘要】：随着人们对近红外光谱分析技术了解的深入,人们发现通过剔除近红外光谱中的冗余变量不仅可以简化近红外光谱分析模型,提高模型的可解读性,通常还可以提高模型的预测效果及稳健性。变量选择的有效性已经在各种近红外光谱应用体系中得到了广泛的验证,发展成为了近红外光谱分析建模过程中一个越来越重要的步骤。为此,化学计量学家们近些年来开发了大量原理不同的新型变量选择算法,基于各种原理的衍生算法也层出不穷。为了让近红外光谱分析研究人员能够较为迅速地对这些算法的特点有所认识,对目前常见的各种变量选择算法的算法原理和优缺点进行了梳理。根据各种算法依据的原理不同,将目前近红外光谱领域常见的变量选择算法大致分为基于偏最小二乘模型参数,基于智能优化算法,基于连续投影策略,基于模型集群分析策略和基于变量区间等五类。在梳理的过程中,我们发现变量选择算法的发展趋势目前主要集中在以下两点:第一,算法的复杂程度不断提高;第二,不同变量选择算法之间的联用开始逐渐增多。此外,作者结合自身在应用变量选择算法时的体会和思考,还总结了变量选择算法在应用层面上存在的一些问题。例如光谱预处理方法对变量选择算法使用效果的影响,以及部分算法存在的稳定性较差,选择变量的可靠性存疑等。
[Abstract]:With the deep understanding of NIR analysis technology, it is found that eliminating redundant variables in NIR spectrum can not only simplify the NIR analysis model, but also improve the interpretation of the model. It can also improve the prediction effect and robustness of the model. The validity of variable selection has been widely validated in various NIR spectrum application systems, and has become an increasingly important step in the modeling process of NIR spectroscopy analysis. Therefore, in recent years, stoichiologists have developed a large number of new variable selection algorithms with different principles, and derivation algorithms based on various principles have emerged in endlessly. In order to enable the researchers to understand the characteristics of these algorithms more quickly, the principles, advantages and disadvantages of the common variable selection algorithms are reviewed. According to the different principles of various algorithms, the variable selection algorithms in the field of near infrared spectroscopy are divided into three parts: partial least squares model parameter, intelligent optimization algorithm and continuous projection strategy. Model-based cluster analysis strategy and variable-based interval and other five categories. In the process of combing, we find that the development trend of variable selection algorithms is mainly focused on the following two points: first, the complexity of the algorithm is increasing; second, the combination of different variable selection algorithms began to increase gradually. In addition, the author also summarizes some problems in the application of the variable selection algorithm based on his own experience and thinking in the application of variable selection algorithm. For example, the effect of spectral preprocessing on the performance of variable selection algorithm, the poor stability of some algorithms, and the doubt of reliability of the selected variable, etc.
【作者单位】：中国农业大学理学院;
【基金】：国家自然科学基金-青年基金项目(31301685)资助
【分类号】：O657.33

【相似文献】