OPTIMIZING EMERGENCY RESPONSE WITH DEEP EMBEDDED CLUSTERING FOR AMBULANCE POSITIONING

Abstract

The number of individuals killed and wounded in traffic accidents is one of the largest problems confronting the contemporary world. Instead of only sending ambulances out when required, pre-positioning them may expedite response times and provide prompt medical treatment. Deep learning techniques hold great potential and have shown to be essential for making decisions and addressing problems in the healthcare sector. This research presents a deep-embedded clustering-based approach to ambulance positing location prediction. Since many patterns and causes within a geographic region have a substantial influence on the frequency of traffic accidents, it is important to comprehend these relationships throughout the model creation process. In order to ensure real-time results, the present study incorporates these patterns using Cat2Vec, another deep learning-based model, and emphasises the need of preserving them during model creation. Furthermore, the proposed framework is compared to traditional clustering methods including GMM, K-means, and Agglomerative Clustering.
In order to evaluate the performance of various algorithms and calculate distance and response time in real time, a special scoring function has also been introduced. The proposed ambulance-positing approach works remarkably well, outperforming all existing traditional methods with a 95% accuracy using k-fold cross-validation and a new distance score of 7.581.