A methodology to predict emergency call high-priority: Case study ECU-911

Abstract

Accurate categorization of emergency calls is essential to determine the most appropriate action towards the solution of an emergency event. There are risks during the caller's attention because the event could be classified incorrectly related to its priority level. In order to reduce the error in classifying those types of calls, a high-priority prediction method of emergencies is proposed. For this, a computational model is presented using text mining techniques, which reduces the high-priority alerts cases wrongly classified as lower-priority alerts. For this, preprocessing techniques were organizing, such as elimination of stop words, lemmatization, and pruning of words according to the frequency in the documents. Inside the validation stage, the Support Vector Machine (SVM) algorithm is proposed, it is focused on confusion matrix optimization in order to reduce cases of false negatives. In other words, the aim is to improve the recall measure in the classification model. The experimentation process revealed that the proposed model improves the prediction of high-priority alerts.