APPROACHES TO BUILDING A HYBRID INTELLIGENT DECISION SUPPORT SYSTEM BASED ON FUZZY LOGIC, XAI, AND LARGE LANGUAGE MODELS

Abstract

. The paper presents the software architecture and practical implementation of an intelligent web-based decision support system that integrates fuzzy logic, explainable artificial intelligence (XAI), and a large language model (LLM) within a unified hybrid framework. A key architectural principle is the separation of the computational decision-making layer from the generative layer. The fuzzy controller processes the input indicators, evaluates them using predefined membership functions and expert rules, and produces a deterministic recommendation. The XAI module then improves transparency by displaying the activated fuzzy rules, membership degrees, and the factors that contributed to the final decision. In contrast, the LLM does not determine the decision itself; instead, it provides context-aware natural-language interpretation, explanatory feedback, and adaptive content based on the computed result. The server side of the system is implemented in Python using the Flask framework, while persistent user and decision data are stored in an SQLite database. Communication with the generative component is organized through an application programming interface. Pilot evaluation demonstrated a high level of agreement with expert judgments, stable system operation, and an end-to-end response cycle suitable for interactive decision support and adaptive web-based applications.