Нейро-нечітка система ANFIS для оцінювання ризику дезінформації в умовах інформаційної війни

Abstract

The article analyses and develops a methodology for using a neuro-fuzzy ANFIS system to assess the risk of disinformation during information warfare. We propose integrating contentbased, network and psychological factors to calculate a single risk indicator (RiskScore) for disinformation spread. The model’s input variables include content characteristics (e.g., truthfulness or emotional tone), network metrics (e.g., propagation speed, reproduction number R₀), and audience psychology (trust level, susceptibility to manipulation, etc.). These variables feed an adaptive fuzzy system that applies a rule-based IF-THEN framework to produce a fuzzy logic risk evaluation. The ANFIS architecture can be trained on limited or synthetic data by tuning rule parameters and membership functions to enhance accuracy. The approach demonstrates an ability to combine heterogeneous factors into a unified metric, enabling early detection of potentially dangerous disinformation campaigns. Combining epidemiological indicators (such as R₀) with contentpsychological features through fuzzy logic improves the informativeness of risk estimates, especially in information-warfare settings. Deploying an ANFIS-based risk-assessment model can help government and defence bodies prioritise counter-disinformation efforts, outperforming traditional linear models, epidemiological SIR models and classical Mamdani fuzzy systems in flexibility. The results provide a foundation for further model development.