Analysis of User Behavioral Patterns on Web Platforms
Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì · 22.06.2026 · 19 · 256-270
DOI: 10.23939/sisn2026.19.256 ↗Kurz gefasst
Fachartikel in der Zeitschrift Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì, online erschienen am 17.05.2026. Verfasst von Dmytro Moroz und 1 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben.
Abstract (Originalsprache)
In the modern digital environment, web analytics and predictive modeling serve as essential tools for optimizing user interaction on Over-the-Top (OTT) platforms. Modern web services increasingly require robust predictive models that not only describe past actions but also forecast future behavior, enabling the delivery of highly personalized content. While structural approaches, such as Markov chains, and classification methods, like Bayesian models, are widely utilized for behavior analysis, their comparative effectiveness within highly imbalanced navigation environments remains insufficiently explored. This creates a critical need to evaluate their respective advantages and limitations for specific business tasks. This study conducts a comprehensive comparative analysis of these probabilistic models using a real-world Netflix clickstream dataset. To ensure analytical precision, the raw data were algorithmically structured into 1,202 autonomous sessions with a 30-minute inactivity threshold. The final empirical dataset comprises 12,104 events generated across 32 unique navigation states, representing the platform's hierarchical structure. The empirical results demonstrate significant differences in how each model handles the specific navigational context of OTT services. The Markov chain model provided superior completeness in identifying target actions within navigation sequences, achieving a Recall of 0.6757, making it highly effective for uncovering implicit user intents. In contrast, the Bayesian model exhibited a lower Recall (0.1199) but achieved a higher overall Accuracy (0.7976 compared to the Markov model's 0.5353). Furthermore, the analysis of the Markov transition matrix revealed the highest transition probabilities between core states. The models also successfully captured prominent behavioral patterns, including "binge-watching". The primary scientific contribution of this paper lies in the quantitative justification of the complementarity of structural and classification approaches. Ultimately, practical testing confirms that integrating Markov and Bayesian approaches allows for a more accurate description of user scenarios, offering actionable insights for UI/UX optimization and the development of personalized recommendation engines.
Zitieren
APA
Moroz, D., & Yuzefovych, R. (2026). Analysis of User Behavioral Patterns on Web Platforms. Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì, 19, 256-270. https://doi.org/10.23939/sisn2026.19.256
Vancouver
Moroz D, Yuzefovych R. Analysis of User Behavioral Patterns on Web Platforms. Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì. 2026;19:256-270. doi:10.23939/sisn2026.19.256
BibTeX
@article{Moroz2026,
author = {Moroz, Dmytro and Yuzefovych, Roman},
title = {Analysis of User Behavioral Patterns on Web Platforms},
journal = {Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì},
year = {2026},
volume = {19},
pages = {256--270},
doi = {10.23939/sisn2026.19.256}
}
EUVIMED