A comparative study on grey relational analysis and C5.0 classification algorithm on adventitious rhizogenesis of Eucalyptus

Abstract

The optimization of plant hormone application to achieve better rhizogenesis in eucalypt cuttings has been demonstrated using a machine learning approach. This study focuses on reducing the human bias in treatment selection in case of adventitious rhizogenesis in eucalypts. The effect of different concentrations of indole-3-butyric acid (IBA) on rhizogenesis was studied and differential responses of genotypes to treatments were observed. Stem cuttings of six different Eucalyptus genotypes were administered with different concentrations and duration of IBA treatments and subsequently multiple parameters, like total length of root system (TLRS), number of roots, shoot to root ratio, etc. were measured using image analysis and manual measurements. A rule-based model and classification tree forming the basis for the C5.0 algorithm was used to eliminate “Human Bias” from the selection procedure. A top-down greedy search approach applied in the whole training set which selected the best feature as the root node and resulted in splitting the data set into smaller subsets. It was subsequently compared with our multiple-attribute decision-making (MADM) model i.e. GRA. The data were trained and tested, and the confidence value report demonstrated a high accuracy level of the decision trees. This suggests the potential use of artificial intelligence in clonal forestry. As per our results, lowering the concentration of auxin and increasing the duration of exposure produced better root quality in E. camaldulensis and hybrids (reciprocal hybrids of E. tereticornis and E. camaldulensis), while the opposite effect was observed in E. tereticornis clones. Therefore, the use of machine learning algorithms could significantly increase the accuracy of treatment selection and provide the optimum results for individual genotypes.