Stetsyk O. Intelligent high-load distributed data processing system in social networks.

Stetsyk O.A. Intelligent high-load distributed data processing system in social networks. Dissertation for the degree of Doctor of Philosophy in the speciality 126 "Information Systems and Technologies." - Kyiv National University of Construction and Architecture. Scientific novelty of the results. The scientific novelty of the dissertation is that: for the first time developed: - a model NMfTCiNF, which is a neural network for text classification in a news feed based on a modification of a hybrid of a convolutional neural network and a bidirectional long short-term memory network, which effectively combines architectural solutions, optimised hyperparameters and pre-trained vector word embeddings using neural networks with transformer architecture in order to classify text efficiently and quickly; - a method for reducing the "bubble" effect through news feed diversification, which is based on the classification of news by the developed neural network into specific discussion categories, and the subsequent clustering of texts in these discussion categories; the proposed method of diversifying the news feed contributes to a more subtle introduction to different points of view regarding a discussion category; improved: - the method of detecting spam and propaganda, by training NMfTCiNF, in the direction of obtaining a balance between accuracy and speed of classification; - the architecture of the social networking system in the direction of bringing the speed of text classifications and generation of a diversified news feed closer to real-time, by combining scalable technologies and trained NMfTCiNF; were further developed: - the concept of creating hybrid systems based on artificial intelligence and machine learning models in the direction of data processing to obtain a cleaner and more diverse environment in social networks.