DATA PROCESSING METHODS FOR ANALYZING PUBLICATION ACTIVITY IN COMPUTER SCIENCE

Badanina Natalya Dmitrievna

The relevance of this study is driven by the fact that over the past decade, computer science has become a field characterized by high-velocity scientific communication. Publication spikes often coincide with the emergence of breakthrough results, such as the release of updated standards or the integration of language models into data analysis practices. However, the scientometric literature lacks standardized tools capable of automatically identifying such short-term anomalies and linking them to external events. The aim of this work is to create a reproducible pipeline for analyzing publication time series in the arXiv computer science category to detect statistically significant spikes and subsequently interpret them by correlating with external events. Methodologically, the research relies on a combination of bibliometric and temporal approaches. The results revealed the presence of two statistically significant spikes. The proposed pipeline enables the quantitative identification and explanation of anomalous surges in publication activity. The findings can be used for the early detection of new research trends and key research groups, which in turn can inform more effective planning of research and development (R&D) directions.

Keywords:computer science, anomaly detection, time series, ARIMA, statistical methods.