Not that kind of tree: Assessing the potential for decision tree-based plant identification using trait databases

Area of Science:

• Botany
• Computational Biology
• Machine Learning

Background:

• Accessible big data and machine learning offer new avenues for enhancing trait-based plant identification.
• The TRY plant trait database provides extensive data for such analyses.

Purpose of the Study:

• To evaluate the efficacy of decision-tree induction for plant identification.
• To identify key plant traits crucial for distinguishing between taxa.

Main Methods:

• Decision trees were constructed using 16 vegetative and floral traits from 689 species across 20 genera.
• The algorithm's classification accuracy was tested on unseen data.
• Trait importance for identification was assessed.

Main Results:

• An unpruned decision tree achieved 98% accuracy in classifying species into genera.
• A pruned tree demonstrated 89% accuracy in classifying new species into genera.
• Seven specific traits were identified as sufficient for classification, providing significant information gain.

Conclusions:

• Tree-based machine learning, leveraging big data, shows significant potential for taxonomic identification.
• Identifying informative traits is key to improving the accuracy and efficiency of plant identification systems.