Natural and non-natural disasters can cause socioeconomic losses and infrastructure damage that hinder sustainable development. Clustering the characteristics of flood impacts across regions can support mitigation planning. However, the classic Fuzzy Geographically Weighted Clustering (FGWC) method can be influenced by the initialization of cluster centers. This study aims to compare the performance of FGWC, FGWC-Flower Pollination Algorithm (FGWC-FPA), and Chaotic Flower Pollination Algorithm–FGWC (CFPA-FGWC) in clustering flood impacts across 38 regencies/cities in East Java using 10 indicators for the year 2025. The evaluation was conducted through 4,400 simulations using eight randomization schemes and six internal validation indices. The results show that CFPA-FGWC with the Circular chaotic map performed best on five of the six indices, while FGWC-FPA produced a slightly lower Classification Entropy value. Two clusters with distinct characteristics emerged. Cluster 1 was dominated by displacement and housing damage, while Cluster 2 exhibited higher frequencies of flooding, fatalities and missing persons, and bridge damage. These findings can serve as a basis for determining aid priorities, evacuation preparedness, and the strengthening of mitigation infrastructure. This study supports SDG 11 and SDG 13 by providing information for disaster risk reduction and enhancing regional resilience.
Copyrights © 2026