Development of a clustering algorithm for parameters of explosive objects based on a comprehensive indicator

Purpose. To enhance the efficiency of clustering parameters of explosive objects through the development of hybrid clustering elements. Methodology. A classifier for explosive objects based on a comprehensive indicator, serving as the main principle for classifier improvement, was developed using mathematical modeling. Data processing was carried out using the Python programming language and scikit-learn libraries. The research methodology involves grouping explosive objects into two clusters with the aim of improving the existing algorithms for detecting explosive objects. Findings. The proposed comprehensive indicator demonstrates a standard deviation 8.2 % less than the existing one. The improved clustering algorithm exhibits Davis-Bouldin index values of 0.517 and 0.525, while the existing ones show 0.572 and 0.572, respectively. This indicates that the output estimations of the new algorithm are less susceptible to noise, which enhances clustering quality and reduces the number of errors during practical application. Originality. A parameter clusterer for explosive objects is proposed which, unlike the existing ones, incorporates complex estimates built on the basis of a linear model with combined parameters as input data. Practical value. The practical significance of the proposed solution lies in the fact that improving existing algorithms for detecting explosive objects will increase the efficiency of computer vision in solving reconnaissance and demining tasks. The proposed solutions can be used as an addition to existing approaches for monitoring and managing national security to prevent emergencies.