@article{M46F11032,
title = "Development of a False Alarm Classification and Prediction Model Using Transformer-Based Large Language Model",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2024",
issn = "1226-4717",
doi = "10.7840/kics.2024.49.4.489",
author = "Jae-hoon Jeong, Hyunho Park",
keywords = "Transformer, NLP, 112, False Alarm, Binary Classification",
abstract = "The 112 emergency reporting system is the police's front line for public safety, and rapid dispatch and incident handling are of the utmost importance. False reports are problematic in that they not only waste police power, but also make it difficult to respond to situations that need help. In order to respond to the increase in false reports, this researcher proposes a 112 false report classification and prediction model based on deep learning. This model receives the report text summarized by the 112 situation room receptionist and determines whether the report is false or misidentified. Mistaken reports are almost impossible to detect on the surface from the point of view of the person who filed the report. Because of this, the researcher conducted an experiment dividing data containing all false reports and data containing only malicious false reports. Model training was performed with the same hyperparameters for five models with transformer structures: BERT,KoBERT, ELECTRA, and RoBERTa. This study is significant in that it took a problem-solving approach to the police's actual security work using natural language processing and LLM. It is expected that the results of this study will help identify malicious false reports and support police decision-making."
}