Development of comprehensive prediction models for pumpkin fruit sensory quality using physicochemical analysis, near-infrared spectroscopy, and machine learning

文献类型： 外文期刊

第一作者： Xu, Yingchao

作者： Xu, Yingchao;Luo, Jiayu;Xue, Shudan;Jin, Qingmin;Zhu, Jitong;Lu, Sen;Meng, Qitao;Du, Hu;Zhong, Yujuan;Fu, Manqin

作者机构：

关键词： Pumpkin; Quality; Spectroscopy; Machine learning; Prediction model

期刊名称：JOURNAL OF FOOD COMPOSITION AND ANALYSIS （ 影响因子：4.0； 五年影响因子：4.1 ）

ISSN： 0889-1575

年卷期： 2024 年 134 卷

页码：

收录情况： SCI

摘要： Pumpkins are vegetables rich in nutrients. Despite the significant impact of sensory attributes on market success, guidance on standardizing quality factors related to sensory quality is lacking. Also, the development of sensory quality prediction models has not been explored in detail. This study aimed to examine 17 quality traits, and proposed evaluation standards incorporating analysis of sensory and key attributes based on 140 cultivars. The correlation analyses revealed significant correlations between taste scores and dry matter, soluble solids, starch, and sucrose contents. A linear equation was established to predict the sensory quality: Y1 (taste scores) = 31.2806 + 0.2011 x X1 (dry matter) + 2.3834 x X2 (soluble solids content) + 0.0917 x X3 (starch) + 0.0307 x X4 (sucrose) - 0.0127 x X5 (malic acid). The credibility of this model, referring to the proportion of variance in the taste scores, was 95%, and the R2 value was 0.841. Meanwhile, regression prediction models for taste score were developed based on near-infrared spectra data using partial least squares and support vector regression, with coefficient of determination (R2) of 1 and 0.895, root mean square errors (RMSEs) of 0.083 and 2.36, R2P of 0.757 and 0.773, and RMSE of prediction (RMSEp) of 3.17 and 3.06 scores, respectively. This study provided a material basis and standard system for evaluating pumpkin quality and highlighted the effective combination of near-infrared spectroscopy with machine learning techniques.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序