EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Kinerja Algoritma PSO Random Forest Versus Metode Konvensional: Imputasi Hujan Harian Jakarta

Ade Suhendar S., Aep Saepuloh

Jurnal Sains dan Teknologi (JSIT) · 18.06.2026 · 6 (2) · 465-475

DOI: 10.47233/jsit.v6i2.4843 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Jurnal Sains dan Teknologi (JSIT), online erschienen am 18.06.2026. Verfasst von Ade Suhendar S. und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Historical daily rainfall data gaps represent an inherent anomaly that fatally distorts the validity of hydrological modeling and disaster early warning systems. This issue is empirically evidenced by an absolute observational failure totaling 6,549 missing days at the Halim Perdanakusuma Meteorological Station over the observation period from 1995 to 2020. This comprehensive study aims to formulate, calibrate, and critically evaluate the effectiveness of an advanced hybrid machine learning framework based on Particle Swarm Optimization - Random Forest (PSO-RF) to rectify these extreme data gap anomalies. Secondary precipitation data from three meteorological stations were systematically partitioned using a rigorous 10-fold cross-validation method to prevent overfitting. The PSO metaheuristic algorithm iteratively explores the hyperparameter space within the Random Forest architecture to minimize predictive deviation. The algorithmic performance of the PSO-RF model is subsequently compared against conventional deterministic spatial interpolation methods, namely Arithmetic Mean and Inverse Distance Weighting (IDW). The statistical significance of the error reduction is rigorously validated using the Wilcoxon Signed-Rank Test. Comparative computational results demonstrate that the PSO-RF hybrid architecture robustly outperforms all conventional methods, achieving the lowest Root Mean Square Error (RMSE) of 16.9593 mm and a positive Coefficient of Determination (R2) of 0.0362. Conversely, conventional methods exhibited severe degradation, yielding an RMSE exceeding 19.9 mm and negative R2 values. Statistical testing confirms that the error reduction by PSO-RF is highly significant (p

Zitieren

APA

S., A. S., & Saepuloh, A. (2026). Kinerja Algoritma PSO Random Forest Versus Metode Konvensional: Imputasi Hujan Harian Jakarta. Jurnal Sains dan Teknologi (JSIT), 6(2), 465-475. https://doi.org/10.47233/jsit.v6i2.4843

Vancouver

S. AS, Saepuloh A. Kinerja Algoritma PSO Random Forest Versus Metode Konvensional: Imputasi Hujan Harian Jakarta. Jurnal Sains dan Teknologi (JSIT). 2026;6(2):465-475. doi:10.47233/jsit.v6i2.4843

BibTeX
@article{S2026,
  author = {S., Ade Suhendar and Saepuloh, Aep},
  title = {Kinerja Algoritma PSO Random Forest Versus Metode Konvensional: Imputasi Hujan Harian Jakarta},
  journal = {Jurnal Sains dan Teknologi (JSIT)},
  year = {2026},
  volume = {6},
  number = {2},
  pages = {465--475},
  doi = {10.47233/jsit.v6i2.4843}
}
Kostenlos starten →