112 신고 건수에 영향을 미치는 주요 변수 분석

Abstract

With the development of data science, various machine learning and deep learning models are being utilized to predict the number of 112 emergency calls with high accuracy. Predicting emergency response volume can help law enforcement agencies efficiently allocate police resources to high-risk areas and dates, thereby preventing crime. However, it is also important to understand the factors that contribute to high call volume on specific dates and times. Police officers may have a sense of whether call volume will be high or low based on their experience and intuition. Similarly, the general public may make assumptions about which factors contribute to higher call volumes. In fact, there is a significant difference in call volume between holidays and weekdays, as several studies have confirmed that holidays have a significant impact on call volume. However, clear variables can reduce the impact of other variables on the prediction model and can become the sole criterion for all predictions. Additionally, variables with relatively low variability may not have a significant impact on call volume prediction. For these reasons, it is important to understand the impact of each variable on prediction results, select the necessary variables for predicting 112 emergency calls, and understand why the model selects certain variables to build an effective prediction model. Through this study, we can examine the key variables that influence the model and their explanatory power, and provide the essential variables needed to construct a 112 emergency call prediction model, without relying solely on model performance improvement.