Predikce vývoje skóre vysokých škol v žebříčku THE World University Rankings pomocí časových řad

This thesis addresses short-term forecasting of university performance in the Times Higher Education World University Rankings for the period 2011 to 2025 and its clear communication through a web application. The workflow follows the CRISP-DM framework covering data acquisition and cleaning, exploratory analysis, model design and evaluation, and deployment of a presentation layer. Forecasts are cast as univariate time series of the overall score and the derived rank. A time-aware backtesting protocol with an expanding window is used to ensure fair model comparison and stepwise error assessment across forecast horizons. Mean absolute error and root mean squared error serve as primary metrics. Emphasis is placed on reproducibility and auditability. Results are stored as versioned artifacts and surfaced in a Streamlit application that supports single-institution forecasts, multi-institution comparisons, and accessible validation summaries. Limitations arise from methodological changes in the ranking and from banded reporting of lower positions. The proposed approach establishes a transferable framework for forecasting in similar ranking settings and for responsible interpretation of results.