Predikcia časových rad

Abstract

This bachelor thesis deals with the issue of time series prediction, which is currently very relevant and important in many areas such as economics, meteorology, or industry. The aim of this work was to describe and compare several different models for time series prediction, which differ in their structure and used methods. Two traditional statistical models, ARIMA and its variations, and the Prophet model were used in this work. In addition to traditional statistical models, the work also focuses on the use of machine learning models, which have become very popular and effective in recent years. Specifically, the Support Vector Regression (SVR) model and LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) neural networks were used. After implementation, the results of predictions of each method were compared and evaluated in terms of their accuracy and efficiency. Overall, this work provides a comprehensive overview of different models for time series prediction and their practical use.