Iqbal Aditya Haqi, Muhammad Ulil Albab, Adi Prasetyo, Nurrofiqi Ankisqiantari

K-NN ALGORITHM AS A DECISION SUPPORT TOOL FOR DETERMINING EVACUATION PRIORITIES IN AREAS AFFECTED BY TSUNAMI POTENTIAL

Introduction

K-nn algorithm as a decision support tool for determining evacuation priorities in areas affected by tsunami potential. Utilize K-NN algorithm as a decision support tool to prioritize tsunami evacuation zones. Assess earthquake tsunami potential & determine critical areas using criteria like distance to sea/high ground.

Abstract

Tsunamis are among the many types of natural disasters that can cause severe and often fatal destruction. A tsunami is defined as a large ocean wave generated by geological activities such as earthquakes, volcanic eruptions, or the shifting of underwater tectonic plates. Due to their high speed and immense power, tsunamis can result in massive flooding and extensive damage to coastal areas. The objective of this study is to assess the tsunami potential of each earthquake event. Furthermore, for those earthquakes identified as having tsunami potential, the study aims to determine which areas should be prioritized for evacuation efforts. This research employs several methods, including literature review, field study, data collection, data analysis, system development, and method validation. The decision-making criteria used in this study include tsunami potential, distance from the sea, distance to high ground, and the number of households. The dataset consists of 30 records, of which 70% are designated for training, while the remaining 30% are set aside for testing purposes. The algorithms applied are K-Nearest Neighbors (K-NN) for classification and the Weighted Product method for decision-making.

Review

This study tackles a critically important and highly relevant topic: leveraging computational tools for tsunami disaster preparedness and evacuation prioritization. The proposed use of the K-Nearest Neighbors (K-NN) algorithm as a decision support tool to determine evacuation priorities in tsunami-prone areas is timely and holds significant practical utility. The objective to assess tsunami potential from earthquake events and subsequently identify priority evacuation zones aligns well with the urgent need for effective disaster management strategies. The application of K-NN for classification, complemented by the Weighted Product method for decision-making, offers a structured approach to a complex, multi-criteria problem, demonstrating a valuable attempt to integrate machine learning into critical humanitarian efforts. While the conceptual framework is promising, the abstract reveals several areas requiring more detailed exposition and potential improvement. The most significant limitation appears to be the extremely small dataset of 30 records, with only 70% (21 records) allocated for training. This size is insufficient for robust K-NN model training and reliable generalization, especially in a high-stakes domain like disaster management where false negatives can have catastrophic consequences. Further clarity is also needed regarding the precise integration of K-NN and the Weighted Product method; specifically, how K-NN classifies tsunami potential and how these outputs feed into the Weighted Product method to determine evacuation priorities. Moreover, the abstract lists decision criteria, but their specific quantification and weighting within the algorithms, along with the details of the "method validation" process, are not sufficiently elaborated. In conclusion, this research presents a commendable initial step towards developing an intelligent decision support system for tsunami evacuation planning. To enhance its scientific rigor and practical applicability, future work should primarily focus on acquiring and incorporating a substantially larger and more diverse dataset. A more thorough discussion of the model's performance metrics, cross-validation techniques, and a sensitivity analysis of the chosen criteria and their weights would also greatly strengthen the study. Exploring the model's scalability, its real-time operational feasibility, and potential integration with geographic information systems (GIS) would further contribute to its real-world impact. Addressing these aspects would transform this promising foundational work into a robust and highly credible tool for disaster preparedness and response.

Full Text

Comments

You need to be logged in to post a comment.