Analýza časových řad

Abstract

Abstract This thesis presents a comprehensive study of time series analysis using deep neural networks. The efficacy of these methods is compared against the existing solutions for time series analysis. The implementation of these methods is done on a selected data set and their performance is verified. The implementation of the methods is done through the use of open source libraries, such as TensorFlow and Keras. Darts framework is used to test existing solutions. The results of the comparison are discussed and conclusions are drawn from them. The overall findings suggest that deep neural networks can be effectively used for time series analysis, with results comparable to existing methods.

Description

Subject(s)

Time Series Analysis, Deep Neural Networks, Time Series Classification, Time Series Prediction, Time Analysis Metrics, RNN, LSTM, GRU, Time Series Framework List, Keras, Darts, KU-HAR Dataset, Yoga Dataset, Chile Wind Speed b08, Jena Dataset, Natural Gas Forecasting

Citation