Návrh a implementace expertního systému pro podporu investičního rozhodování podle Clean Architecture

Abstract

Investment decision-making in financial markets is characterized by a high degree of uncertainty, complexity, and dependence on a large amount of heterogeneous information. This process is often influenced by the investor’s subjective judgment, which may lead to suboptimal decisions. The use of expert systems enables the formalization of knowledge and decision-making procedures and their consistent application in the evaluation of investment opportunities.

The aim of this thesis is to design and implement an expert system for investment decision support that, based on explicitly defined rules and knowledge, enables systematic evaluation of stocks and provides investment recommendations. The system design is based on the principles of rule-based expert systems and is implemented using the Clean Architecture approach.

As part of the solution, an investment questionnaire was developed to classify users into investor profiles, and a knowledge base containing rules based on financial metrics of stocks was designed. The inference mechanism subsequently evaluates the fulfillment of these rules and, using a scoring approach, generates the final recommendation.

The result of the thesis is a functional expert system that enables the analysis of stocks based on defined rules and adapts recommendations to individual user preferences through investor profiles and configurable threshold values. The system contributes to reducing subjectivity in decision-making and provides a structured view of investment opportunities.

The proposed solution demonstrates the potential of rule-based expert systems in the field of investment decision support and confirms the suitability of applying Clean Architecture principles in the design of such a system. A limitation of the approach is primarily its dependence on statically defined rules and the quality of input data, while system flexibility is partially ensured through configurable threshold values.