Development of intelligent analysis and forecasting systems based on machine learning: an integrated approach to solving problems of materials science and cybersecurity
Keywords:
machine learning, deep learning, composite materials, bot detection, graph neural networks, transformers, prediction of material properties, social network analysis, intelligent systems, artificial intelligence.Abstract
The paper proposes an integrated approach to the development of intelligent analysis and forecasting systems using modern machine learning methods. The research covers two areas: predicting the properties of composite materials and detecting bots in social networks. To solve the first problem, a hybrid model was developed that combines physical principles and deep learning methods, providing high accuracy in predicting the mechanical characteristics of materials. To detect bots, a multimodal approach was applied, combining text analysis with pre-trained language models with graph analysis of social structures, which allowed achieving indicators superior to existing solutions. The experimental verification was performed on large-scale datasets, including more than 10 thousand samples of materials and 1 million profiles of users of social networks. The results obtained demonstrate an improvement in the quality of forecasting and classification compared to current methods. The practical significance of the work lies in the creation of scalable systems suitable for use in designing new materials and monitoring online communities.
