Can you make morphometrics work when you know the right answer? Pick and mix approaches for apple identification.
Can you make morphometrics work when you know the right answer? Pick and mix approaches for apple identification.
复制标题
DOI：
10.1371/journal.pone.0205357
复制
发表时间：
2018
期刊：
影响因子：
3.7
通讯作者：
Culham A
中科院分区：
文献类型：
--
作者：
Christodoulou MD;Battey NH;Culham A
几个世纪以来，生物的形态分类一直是科学的挑战，并在数据收集和分析方面产生了广泛的客观形态测量方法。在本文中，我们用苹果品种来探索这些方法，苹果品种是一个模型生物系统，其中离散的群体是预先定义的，但在整体形态相似性方面有很高的水平。使用统计学习工具评估形态测量技术在发现组中的有效性。没有一种技术在每次分类中都被证明是最优的，线性形态测量技术的表现略优于几何技术（在测试集上的准确率为72.6%对66.7%）。将这些技术与他们对特定品种的个别成功的事后知识相结合，可以获得显著更高的分类准确率（77.8%）。由此我们得出结论，即使有预先确定的离散类别，也需要一系列方法，其中这些类别本质上彼此相似，并且我们提出了一个问题，即在分类潜在连续自然变化的研究中，类别之间的匹配水平是否通常设置得太高。
Morphological classification of living things has challenged science for several centuries and has led to a wide range of objective morphometric approaches in data gathering and analysis. In this paper we explore those methods using apple cultivars, a model biological system in which discrete groups are pre-defined but in which there is a high level of overall morphological similarity. The effectiveness of morphometric techniques in discovering the groups is evaluated using statistical learning tools. No one technique proved optimal in classification on every occasion, linear morphometric techniques slightly out-performing geometric (72.6% accuracy on test set versus 66.7%). The combined use of these techniques with post-hoc knowledge of their individual successes with particular cultivars achieves a notably higher classification accuracy (77.8%). From this we conclude that even with pre-determined discrete categories, a range of approaches is needed where those categories are intrinsically similar to each other, and we raise the question of whether in studies where potentially continuous natural variation is being categorised the level of match between categories is routinely set too high.
登录
查看更多内容
影响因子：
--
作者：
Adams, DC;Rohlf, FJ;Slice, DE
通讯作者：
Slice, DE
影响因子：
3.7
作者：
Corney, David P. A.;Tang, H. Lilian;Jin, Jing
通讯作者：
Jin, Jing
影响因子：
7.5
作者：
CORTES, C;VAPNIK, V
通讯作者：
VAPNIK, V
影响因子：
2.4
作者：
Guisande, C.;Manjarres-Hernandez, A.;Salmeron, F.
通讯作者：
Salmeron, F.
影响因子：
3.4
作者：
Corney, David P. A.;Clark, Jonathan Y.;Wilkin, Paul
通讯作者：
Wilkin, Paul