IMPLEMENTATION OF UNSUPERVISED LEARNING MODELS FOR ANALYZING THE STATE'S SECURITY LEVEL

Objective. Enhancing the effectiveness of preliminary analysis of the state's security level through the implementation of clustering models. Methodology. The process of creating unsupervised learning models and their peculiarities in tasks of analyzing the state's security level has been investigated. Techniques for creating the basic k-means model and its improvement through the use of Pearson correlation as a distance metric have been considered. Determining cluster centers was performed by both the basic method and the Cochran's maps method. The optimal quality indicator, according to the results of clustering, was considered to be the model demonstrating the minimum value of the Davies-Bouldin index. Results. An improved unsupervised learning model based on the k-means algorithm for analyzing the state's security level has been developed. The model is characterized by two clusters, with centroids determined as 1.112 and 1.009. Scientific novelty. The proposed model for clustering the state's security level differs from existing ones by using as input estimates derived from a comprehensive indicator based on the principles of interaction and emergent properties. This allows obtaining advantages of the clustering model in terms of the Davies-Bouldin index. The existing clustering model demonstrates a value of 0.4765, while the proposed one achieves 0.2166. Practical significance. The proposals serve as a useful additional tool for preliminary analysis of the state's security level during air alerts and extend the functionality of the previously researched forecasting technology.