Результати пошуку

Зараз показуємо 1 - 3 з 3

Effective Content Moderation Using Modern AI Tools
(2024) Bortnyk, Kateryna; Bahniuk, Nataliia; Kondius, Inna; Melnyk, Kateryna; Melnychuk, Yuliia; Kondius, Kostiantyn
In the paper we develop a web application that integrates AI capabilities for content moderation, while providing an effective and balanced user protection system based on general comparative analysis of two leading tools for content moderation: Microsoft Content Moderator and Google Perspective API. Moderation mechanisms using AI algorithms have been developed, allowing not only to filter unwanted content, but also to adapt to changing conditions and user needs. We present the comparison results of two language models of Google Perspective API and Microsoft Content Moderator and appropriate tools taking into account the specific helpful, honest, harmless criteriums that is aligned with human values and include emotional tonality supported by developers.
The system of dynamic optimization pricing by machine learning
(Greece, 2024) Konotopchyk, Artem; Melnyk, Kateryna; Lavrenchuk, Svitlana; Khrystynets, Nataliia; Melnyk, Pavlo; Melnyk, Pavlo; Bortnyk, Kateryna
Based on the real retail data collected over eight months, a number of models are developed to determine the most efficient machine learning algorithm. It is found that the KNeighbors Regressor model demonstrates the best performance for a small number of transactions, achieving a low MSE of 0.00091 and a high R2 of 0.72 in the validation set. For a large number of transactions, Random Forest Regressor and Decision Tree Regressor show the best ability to capture complex relationships and handle nonlinearties in the data, thanks to the ensemble learning technique. Whereas the Linear Regressor and Support Vector Regressor models demonstrated large deviations from the real price on the test data. The results of the study demonstrate the relevance of rtificial intelligence algorithms in dynamic pricing trategies, showing the ability to quickly process huge amounts of data, taking into account numerous factors and changes in market demand.
Threats detection and analysis based on SYSMON tool
(2023) Bahniuk, Nataliia; Linchuk, Oleksandr; Bortnyk, Kateryna; Kondius, Inna; Melnyk, Kateryna; Kondiu, Kostiantyn
In this work, an nalysis for the study of threats in a real environment with the possibility of conducting a fullfledged analysis of threats, as well as their simulationhas been developed for research purposes. Designed laboratory was built for the threats research, specification of deploying and configuring Sysmon, imitation of an attack in laboratory conditions and its investigation by implicit signs, the processesing of threat analysis using the Sysmon tool. We present a system based on the analysis of continuous input chan-nels of Sysmon logs. The system is based on the Cyber Threat Analysis Ontology and analyzes SYSMON logs to classify software according to different threat levels and enhance cyber defense capabilities with situational awareness, prediction and auto-mated actions. The developed laboratory improves the effectiveness of threat analysis using the Sysmon tool, makes study of threats, deploying and configuring Sysmon, imitation of an attack in laboratory conditions and its investigation by implicit signs. It can be applied for the study of threats in a real environment with the possibility of conducting a full-fledged analysis of threats, as well as their simulation for research purposes.

Наукові статті

Переглянути

Фільтри

Налаштування

Сортувати за

Результатів на сторінку

Результати пошуку