Improved version of explainable decision forest: Forest-Based Tree | ||||
IJCI. International Journal of Computers and Information | ||||
Article 5, Volume 10, Issue 1, January 2023, Page 54-64 PDF (957.6 K) | ||||
Document Type: Original Article | ||||
DOI: 10.21608/ijci.2022.155977.1082 | ||||
![]() | ||||
Authors | ||||
Faten Khalifa ![]() | ||||
1Information Systems dept., Faculty of Computers and Information, Menoufia University, Shebin Elkom 32511, Menoufia, Egypt | ||||
2Information System, faculty of computer and information, Menoufia University, Shebin El Kom, Menofia, Egypt | ||||
3Information SystemsDepartment Faculty of Computers and Information Menoufia University, Egypt | ||||
Abstract | ||||
A Decision Forest is an ensemble learning method that seeks to enhance the predictivity of a single decision tree via training several trees and combining their decisions. However, it is not easy to explain the rationale behind the predictions of decision forests; as each prediction consists of an integration of many decisions. This defect of decision forest makes it a black box, missing interpretable capability, and it will be difficult for humans to understand its entire logic. In this article, we discuss the transformation of the decision forest into a single decision tree, Forest-Based Tree (FBT), without sacrificing accuracy. The proposed method combines the decision rules of individual trees and organizes them into a tree structure. We focus on how to optimize the algorithm; to build an intelligible and lightweight forest-based tree quickly. The open source software for FBT is also provided. Results on 30 UCI datasets show the objective to approximate the predictive performance of the decision forest through forming an explainable decision tree in less computational time. | ||||
Keywords | ||||
Decision forest; ensemble pruning; ensemble-derived models; explainable AI; random forest | ||||
Statistics Article View: 287 PDF Download: 304 |
||||