On the Interpretability of a Tree-based Ensemble for Predicting Software Effort

Precisely predicting software development effort is crucial for effective project monitoring and management. While machine learning-based techniques are more accurate than classical methods, professionals are hesitant to use them due to the lack of intuition and explanation in generated estimates. T...

Full description

Saved in:
Bibliographic Details
Main Authors: Najm, Assia, Zakrani, Abdelali
Format: Conference Proceeding
Language:English
Subjects:
c-fuzzy regression tree
Estimation
Information science
interpretability
Machine learning
Monitoring
optimal tree
Predictive models
Regression tree analysis
Software
software effort estimation
software effort interpretation
tree ensemble
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Precisely predicting software development effort is crucial for effective project monitoring and management. While machine learning-based techniques are more accurate than classical methods, professionals are hesitant to use them due to the lack of intuition and explanation in generated estimates. This study addresses this gap by examining the interpretability of machine learning models in software effort prediction. Using the ensemble of optimal additive cluster-based fuzzy regression trees (Opt-ACFRT) model, estimations are performed on two datasets (ISBSG, Albrecht) using the 30% holdout cross-validation method. Global and local interpretability methods are employed to enhance trust and comprehension of the estimates. The results demonstrate that these techniques effectively assist software managers and practitioners in obtaining understandable estimates and making informed decisions.
ISSN:2327-1884
DOI:10.1109/CiSt56084.2023.10410020