Claim Missing Document
Check
Articles

Found 1 Documents
Search

D Decision Tree-Based Classification of Guimaras Mango Varieties Using Fruit and Stone Attributes with the J48 Algorithm Rifqi Faiza Mohammad; Fajar Delli Wihartiko; Gustian Rahmaputra; Adrian J. Forca
JURIKOM (Jurnal Riset Komputer) Vol. 13 No. 5 (2026): Oktober 2026
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/jurikom.v15i5.10156

Abstract

Guimaras mango varieties have overlapping physical characteristics that complicate manual identification. This exploratory study compares three J48 decision-tree configurations using 11 recorded observations (one per variety) and 17 candidate predictors describing fruit and stone attributes in WEKA. Model A uses pruning with a confidence factor of 0.25 and a minimum leaf size of two; Model B uses the same confidence factor with a minimum leaf size of one; Model C is unpruned with a minimum leaf size of one. The trees contain 7, 19, and 22 nodes, with 5, 13, and 15 leaves, respectively. Peel color is the root attribute in all three configurations; subsequent splits also use fruit weight, flesh color, peel thickness, shape, and, in Model C, peel texture. Evaluation on the training observations produces 45.45% accuracy (5/11) for Model A and 100% (11/11) for Models B and C. These are resubstitution results, not validated estimates of performance on unseen specimens. Stone predictors were available but were not selected in the fitted trees. This study demonstrates interpretable, parameter-sensitive descriptions of the recorded classes; independently replicated fruit samples are needed before generalization or practical varietal authentication can be established.