Building a model for predicting fire dynamics

Department of Business Informatics and Business Process Modeling, Siberian Federal University, 79 Svobodny Ave., Krasnoyarsk, 660041, Russia

^* Corresponding author: lev_kene@mail.ru

Abstract

The paper compares known classification algorithms (logistic regression, k-nearest neighbors, support vector machine, stacking) on the problem of predicting the fire class. A feature of this problem is that it takes into account the specifics of the initial data at the exploratory analysis stage, i.e. before using the initial data by classification algorithms. First, at the exploratory analysis stage, it is necessary to solve the problem of selecting factors to predict the fire class. Second, check the initial data for gaps and outliers. Third, normalize the data in order to prepare them for building a model for predicting fire dynamics. Experimental studies are conducted on data on fires in the Krasnoyarsk Krai from 2010 to 2020.