Numerical method for modification of models developed based on the method of hierarchy analysis, using an artificial neural network

Abstract

The article considers the process of forming models for assessing the quality indicator using the hierarchy analysis method in subject areas characterized by related features. The use of the classical hierarchy analysis method does not allow forming models with related features. The purpose of the work is to develop and verify a numerical method that will allow modifying models formed using the hierarchy analysis method and increasing their accuracy by taking into account the relationship of features. Verification of the developed method is carried out on the example of forming models for assessing the danger of destructive software impacts of malware of the malicious utilities class on automated special-purpose systems of internal affairs agencies. Based on the hierarchy analysis method, based on the results of a survey of experts in the field of information security, the initial model is formed. Related behavioral patterns of malicious utilities are revealed, the joint implementation of which leads to an increase in the value of the danger indicator of the alternative under study. To study the related features, an artificial neural network of direct propagation is formed. The hyperparameters of the artificial neural network are defined in such a way as to obtain sufficient accuracy during verification. A numerical method has been developed that allows taking into account related features in models developed using the hierarchy analysis method. The numerical method was used to modify and verify the original model. A test data set generated during an expert survey was used to verify the model. Verification of the numerical method showed the consistency of the results obtained. The results of the study can be used by specialists to form models for assessing the quality indicator with related features.