Scholarlink

<p>Near-infrared (NIR) spectroscopy plays a critical role in quality assessments for pharmaceuticals and food industries. Predictive model accuracy is substantially improved through optimized methods of pre-processing complex NIR data. Researchers optimize Partial Least Squares models through techniques like normalization, baseline correction, and wavelength selection to enhance data quality, reduce noise levels, and improve performance. This research conducts a systematic literature review (2019-2024) to identify research gaps and thematic connections by analyzing publications from 15 major scientific databases. The initial search resulted in screening 21,850 papers before reaching <b>47</b> papers that met the selection criteria. Analysis of the 47 studies revealed that preprocessing effectiveness is highly dataset-dependent, with no single technique universally optimal across applications. The evaluation of existing research uses structured research questions to help practitioners and researchers create an optimal methodology that links pre-processing methods with Partial Least Squares (PLS) models, thereby improving model accuracy and data quality across various applications. The review identifies existing analytical problems and proposes meaningful solutions for these issues. Research findings provide guidance to professionals as well as scientists to develop new innovations that enhance PLS model data quality and predictive performance through optimized approaches. Standardized NIR analysis methods require further research attention to fill gaps while providing predictive effectiveness along with adaptable use in diverse applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing NIR spectral preprocessing for PLS predictive accuracy: a systematic review (2019-2024)

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月25日

查看全文
  • Malayka Wahab,
  • Ashraf Zia,
  • Waqas Ahmad,
  • Muhammad Zakarya,
  • Muhammad Haleem

摘要

Near-infrared (NIR) spectroscopy plays a critical role in quality assessments for pharmaceuticals and food industries. Predictive model accuracy is substantially improved through optimized methods of pre-processing complex NIR data. Researchers optimize Partial Least Squares models through techniques like normalization, baseline correction, and wavelength selection to enhance data quality, reduce noise levels, and improve performance. This research conducts a systematic literature review (2019-2024) to identify research gaps and thematic connections by analyzing publications from 15 major scientific databases. The initial search resulted in screening 21,850 papers before reaching 47 papers that met the selection criteria. Analysis of the 47 studies revealed that preprocessing effectiveness is highly dataset-dependent, with no single technique universally optimal across applications. The evaluation of existing research uses structured research questions to help practitioners and researchers create an optimal methodology that links pre-processing methods with Partial Least Squares (PLS) models, thereby improving model accuracy and data quality across various applications. The review identifies existing analytical problems and proposes meaningful solutions for these issues. Research findings provide guidance to professionals as well as scientists to develop new innovations that enhance PLS model data quality and predictive performance through optimized approaches. Standardized NIR analysis methods require further research attention to fill gaps while providing predictive effectiveness along with adaptable use in diverse applications.