METHODS OF VISUAL, TEXTUAL, AND WEB DATA ANALYSIS IN MODERN DATA MINING SYSTEMS

Abstract

The article examines a comprehensive approach to data analysis in modern Data Mining systems, which integrates methods of textual, visual, and Internet-oriented information processing. It is shown that the heterogeneity of contemporary data, encompassing textual, behavioral, structured, and dynamic information streams, necessitates the use of multi-level analytical technologies capable of operating simultaneously with different types of representations and models. A review of scientific sources was conducted, highlighting the development trends of natural language processing models (including deep learning techniques), visual analytics tools, and web mining methods, as well as outlining their advantages and limitations in practical applications. The study substantiates the feasibility of integrating these three groups of methods into a unified analytical platform that ensures deep semantic processing of data, enhanced interpretability of results, and the ability to handle highly dynamic Internet data flows. To evaluate the effectiveness of the proposed approach, a prototype system was developed and tested on an online-catalog dataset that included user textual reviews, behavioral activity logs, and structured product attributes. The experimental results demonstrate that combining textual features with behavioral web indicators improves classification performance compared to relying solely on textual data. The integration of visual analytics, in turn, significantly reduces the time required for interpreting the results and enhances user experience, which is confirmed both by operational metrics and expert evaluations. The findings indicate that the integrated approach is promising for developing scalable, adaptive, and interpretable Data Mining systems designed for real-world big data processing conditions. Future research should focus on advancing universal models capable of combining textual, behavioral, graph-based, and spatiotemporal data, as well as improving explainable artificial intelligence tools for analyzing complex analytical processes.