|
|
Time-Series Analysis and Prediction by Means of Neural Networks
Kňažovič, Martin ; Jaroš, Jiří (referee) ; Bidlo, Michal (advisor)
This thesis deals with stock price prediction based on the creation of prediction models for selected stocks (BRK-A, GOOG, and MSFT), which can help investors in the creation of their financial decisions or by replacing other stock prediction models in existing prediction systems. Models created in this thesis are presented in two types - univariate model and multivariate model, which are in their final version presented in two architectures, one-layer architecture and two-layer architecture. Discussed models are created by means of neural networks, specifically recurrent neural networks with its extension - Long short-term memory. The output of the presented models is a forecast of the next-day stock price, which can be used for evaluating the right time to buy or sell a given stock. The quality of individual prediction models is evaluated via the mean squared error of the validation or testing dataset or alternatively based on stock price trend prediction.
|
|
|
Techniques For Avoiding Model Overfitting On Small Dataset
Kratochvila, Lukas
Building a deep learning model based on small dataset is difficult, even impossible. Toavoiding overfitting, we must constrain model, which we train. Techniques as data augmentation,regularization or data normalization could be crucial. We have created a benchmark with a simpleCNN image classifier in order to find the best techniques. As a result, we compare different types ofdata augmentation and weights regularization and data normalization on a small dataset.
|
|
|
Composite indicators: the construction, usage and interpretation
Hudrlíková, Lenka ; Fischer, Jakub (advisor) ; Čadil, Jan (referee) ; Hužvár, Miroslav (referee)
This thesis brings a comprehensive view on the construction, usage and interpretation of composite indicators. Methods and techniques, which can be used for constructing composite indicators, are introduced. The focus is on their contribution to the transparent solution of the problem of correlation and compensability among underlying indicators. Transparency in construction of composite indicators is a crucial requirement for obtaining reliable results and their correct interpretation. The thesis consists of two main parts. The first part is theoretically oriented. First, the problem of adequacy and subsequently a measurement of the phenomenon by means of statistical indicators are discussed. Different methods for data normalization, setting a weighting scheme and aggregation are introduced and compared. These three steps are considered to be crucial in a process of constructing a composite indicator and thus, they are the core of the thesis. The aim is to investigate an interaction of normalization methods, weight-setting and aggregation methods, since these steps are not separate. The second part of the thesis consists of two comprehensive cases. Theoretical findings are applied and empirically verified in these cases. I investigated a robustness of the composite indicator depending on a combination of selected methods of normalization, setting weights and aggregation on a set of Europe 2020 indicators. Whereas this first case dealt with the comparative analysis of methods, the second case is focused purely on one issue -- university ranking. The proposed method reacts to criticism of currently published university rankings and takes into account specifics of the particular university as well as the exogenous background characteristics. The main added value rests in a contribution to a discussion about the improvement of construction and overall quality of composite indicators including their interpretation. I pointed out the main concerns and difficulties of composite indicators that often remain unnoticed by users and even constructors. The conclusion brings several beneficial findings, which can be used for the construction of a composite indicator and an interpretation of final scores and ranking. This work can also serve as a scientific ground for further research and development of the methodology of constructing composite indicators.
|
guest ::
