Research on Artificial Intelligence Models for Predicting User Behaviour in Digital Systems
Students Name: Pshenychka Vasylyna Vasylivna
Qualification Level: magister
Speciality: System Administration of Telecommunications Networks
Institute: Institute of Information and Communication Technologies and Electronic Engineering
Mode of Study: full
Academic Year: 2025-2026 н.р.
Language of Defence: ukrainian
Abstract: The master’s thesis is devoted to the research, development, and experimental validation of an artificial intelligence model designed to predict user behavior in digital systems. With the ongoing digital transformation of modern society, user behavior has become a key success factor for information services, marketing platforms, and data analytics systems. At the same time, the dynamic and unpredictable nature of behavioral processes requires the use of advanced intelligent tools capable of automatically identifying patterns, predicting future user actions, and supporting managerial decision- making. The aim of the research is to develop an adaptive artificial intelligence model based on user activity data that can accurately forecast behavior and ensure efficient allocation of digital interactions. The first chapter provides a comprehensive theoretical analysis of user behavior in digital environments, emphasizing the complexity of this phenomenon, which integrates technological, social, psychological, and informational factors. Using a structured approach, the study describes the stages of behavioral model formation, mechanisms of user interaction with information systems, and classification of responses to digital stimuli. Moreover, it highlights the role of digital platforms as sources of behavioral data and real- time mechanisms that form the basis for predictive modeling. The chapter also focuses on ethical, security, and transparency issues related to the use of personal information, which are crucial for the development of modern artificial intelligence systems [1–3]. The second chapter analyzes the theoretical and practical aspects of artificial intelligence development, its classification, and key methodologies. It provides a detailed examination of the principles of classification algorithms, ensemble methods, neural networks, and support vector machines applied to the prediction of complex behavioral models. The advantages and limitations of each method are discussed, along with their applicability to large datasets and dynamic behavioral processes. The chapter emphasizes the importance of ensuring interpretability of model outputs and the responsible use of AI technologies in domains that involve personal user data. The third chapter addresses the formulation of the user behavior prediction problem and presents a detailed description of the model construction process, including data preparation, feature selection, training, and parameter optimization. The methodological framework includes defining the target function, determining behavioral attributes, and establishing evaluation metrics to assess predictive accuracy. The fourth chapter presents the results of the experimental evaluation of the developed model. Three machine learning algorithms were tested: Random Forest, Support Vector Machine, and Artificial Neural Network. Comparative analysis demonstrated that the Random Forest and Neural Network models achieved the highest accuracy, confirming the efficiency of a hybrid approach to user behavior prediction. Feature importance analysis revealed that user response history, activity level, and interaction frequency were the most influential predictors. A comprehensive temporal analysis identified stable behavioral patterns that enabled the segmentation of users with distinct temporal preferences (morning and evening activity). Independent testing confirmed the persistence of these behavioral patterns and demonstrated the potential of using temporal features as a foundational tool for personalized digital strategies. The developed model exhibited high predictive accuracy, adaptability, and cost-efficiency, showing that personalized timing and content of communication can reduce information expenses and enhance marketing effectiveness [4–9]. The conducted research expands the scientific understanding of user behavior mechanisms within digital ecosystems by integrating methods from sociology, informatics, and mathematics. From a practical standpoint, it proposes a comprehensive approach to building adaptive predictive models that can be implemented in domains such as e-commerce, digital marketing, online education, and user experience management. Overall, the results demonstrate that the developed artificial intelligence model serves as an effective instrument for predicting user behavior and supporting decision-making in digital systems. It maintains a balance between accuracy, interpretability, and ethical use of data. Future research may focus on enhancing hybrid modeling approaches, incorporating contextual factors, and developing real-time adaptive systems for dynamic prediction and interaction optimization. Study object - User behavior in digital systems. Scope of research - Application of artificial intelligence for analyzing and predicting user behavior in digital environments. Goal of research - Investigation of an artificial intelligence model with the aim of predicting user behavior in digital systems and supporting effective managerial decisions. The study examines the design, implementation, and validation of an artificial intelligence model for predicting user behavior in digital systems. Integrating machine learning, behavioral analytics, and temporal pattern recognition, the model demonstrates high predictive accuracy and adaptability. It enhances personalization, optimizes user interaction, and supports data-driven decision-making in modern digital ecosystems. The results confirm its effectiveness for intelligent automation and behavioral forecasting.