Investigating Tree-Based Classifiers and Selected Ensemble Learning on Iris Flower Species Classification

- Eloquence, hope, knowledge, the ability to communicate effectively, and faith are some of the meanings associated with the iris flower in the language of flowers. Iris has different species types, and each type has its own medicinal purpose. Classifying the flower has become a serious task for researchers due to the high volume of datasets (big data), hence the introduction of machine learning algorithms for accurate and reliable classification. This paper focuses on the classification of the Iris flower using five tree-based algorithms; Best First Tree (BFTree), Least Absolute deviation Tree (LADTree), Cost-Sensitive Decision Forest (CSForest), Functional Tree (FT) and Random Tree (RT). Three selected ensemble learning (Bagging, Dagging and cascade generalisation) were equally implemented in the algorithm. The dataset that was utilised in this investigation is open source and may be downloaded without cost from a public repository (kaggle.com). The result of the classification showed that the FT classifiers outperform other tree-based classifiers with an accuracy of 96.67% and an AUC of 1.00. The ensemble algorithm has a significant impact on the performance of single classifiers (tree-based). Outperform tree based. AUC/ROC (Area Under Curve/Receiver Operating Characteristics) was used to evaluate the algorithm's performance. Bagging ensemble outperforms other ensembles (Dagging and Cascade) with an accuracy of 96.00% and AUC of 1.00.