Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2025
- Goals
- Record
- 23321449-88f4-4516-91f9-cf34d3e7c335
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), with Winarsih, Nurul Anisa Sri; Pramunendar, Ricardus Anggi; Shidik, Guruh Fajar; Widjajanto, Budi; Rohman, Muhammad Syaifur; and Ratmana, Danny Oka as participants, conducted research on optimizing earthquake damage prediction through particle swarm optimization-based feature selection. The research addressed the need for accurate earthquake damage predictions because emergency response actions immediately after an earthquake can reduce economic losses and save lives.
The activity examined machine learning techniques for predicting earthquake damage. The algorithms used were k-nearest neighbors, decision tree, random forest, and Naive Bayes. The research also examined feature selection, which selects the most relevant features from big data. Particle swarm optimization was used for machine learning optimization and feature selection. Various particle swarm optimization approaches were compared.
Random forest with Phasor particle swarm optimization achieved the highest fitness score. The process reduced the features from 38 to 14. After feature selection, the k-nearest neighbors, decision tree, and random forest algorithms improved. Random forest achieved the best accuracy at 72.989%. Processing time in the decision tree, random forest, and Naive Bayes algorithms was faster than before. The research concluded that combining machine learning algorithms with particle swarm optimization feature selection produced a classification model with better performance than a model without feature selection.
This UDINUS research provides documented evidence of work on earthquake damage prediction, feature selection, classification, and processing performance.