K-Nearest-Neighbours-Based Fuzzy Regression for Modelling Heterogeneous Strength Development in Cement-Based Materials
Journal of Science and Mathematics Letters · 25.07.2026 · 14 (3) · 475
DOI: 10.37134/jsml.vol14.3.9.2026 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Science and Mathematics Letters, online erschienen am 25.07.2026. Verfasst von Hazmira Yozza. Die Arbeit nennt 48 Literaturangaben.
Abstract (Originalsprache)
Compressive strength is a key indicator of cement performance. The 28-day compressive strength is a benchmark for compressive strength development. In most studies, compressive strength data are treated as precise values. However, compressive strength measurements inherently involve uncertainty due to experimental variability, material heterogeneity, and testing conditions. This uncertainty suggests that compressive strength may be appropriately represented as a fuzzy number and analysed using fuzzy regression. This study aims to investigate the use of integrated k-nearest-neighbour fuzzy regression (I-KNNFR) to predict the triangular fuzzy number of the 28-day compressive strength of cement mortar. For each prediction point, the method constructs a possibilistic fuzzy regression model using locally similar observations identified through the nearest-neighbour mechanism. The I-KNNFR was developed to address issues in possibilistic fuzzy regression related to global modelling, outliers, and over-constrained problems. Its predictive performance was evaluated using 10-fold range-preserved controlled cross-validation and compared with that of possibilistic fuzzy regression based on mean squared error (MSE) and mean relative distance (MRD). The best predictive performance was achieved when I-KNNFR used nine nearest neighbours and the Euclidean distance. Compared with possibilistic fuzzy regression, I-KNNFR reduced MSE and MRD by approximately 85% and 83%, respectively, indicating a substantial improvement in predictive performance. In conclusion, I-KNNFR more effectively captures local strength-development patterns through fuzzy modelling. It outperforms possibilistic fuzzy regression in predicting the triangular fuzzy number of the compressive strength. By using locally selected neighbours, it also represents the uncertainty associated with 28-day compressive-strength prediction.
Zitieren
APA
Yozza, H. (2026). K-Nearest-Neighbours-Based Fuzzy Regression for Modelling Heterogeneous Strength Development in Cement-Based Materials. Journal of Science and Mathematics Letters, 14(3), 475. https://doi.org/10.37134/jsml.vol14.3.9.2026
Vancouver
Yozza H. K-Nearest-Neighbours-Based Fuzzy Regression for Modelling Heterogeneous Strength Development in Cement-Based Materials. Journal of Science and Mathematics Letters. 2026;14(3):475. doi:10.37134/jsml.vol14.3.9.2026
BibTeX
@article{Yozza2026,
author = {Yozza, Hazmira},
title = {K-Nearest-Neighbours-Based Fuzzy Regression for Modelling Heterogeneous Strength Development in Cement-Based Materials},
journal = {Journal of Science and Mathematics Letters},
year = {2026},
volume = {14},
number = {3},
pages = {475},
doi = {10.37134/jsml.vol14.3.9.2026}
}
EUVIMED