|
|
The use of convolutional neural networks for predicting the financial failure of a company
Šebestová, Monika ; Chramcov, Bronislav (referee) ; Lenort, Radim (referee) ; Režňáková, Mária (referee) ; Dostál, Petr (advisor)
The doctoral thesis deals with the use of convolutional neural networks for predicting the financial failure of companies. A bibliometric analysis was used during the processing of the literature review, which enabled a better orientation in scientific works oriented to the methods and approaches used in the past to predict the financial failure of companies. On the basis of the obtained knowledge, a deep learning model based on the GoogLeNet architecture was proposed, with inputs consisting of financial and macroeconomic indicators of companies. The modeling was based on the transfer learning method, in which it is possible to fine-tune the parameters of the pre-established networks to accelerate the learning process of the convolutional neural network. The initial set of financial and macroeconomic indicators was compiled from the variables that were most often used in business failure prediction models in scientific papers. Appropriate statistical methods were used for the specific selection of indicators from which the model is built. Since convolutional neural networks work best with image processing, the quantitative values of the input indicators were graphically interpreted and it was investigated which type of graphical processing is most suitable for predicting the failure of companies. Due to the existence of an unbalanced data set, the effect of the SMOTE method on the accuracy of the model's prediction was analyzed in the thesis. The method was used to increase the number of samples of the minority class of firms. To model the prediction of financial default, several variants of models were proposed, which differed in the form of input data. It was tested how the removal of outliers from the data set, the point in time from which the data come or the method of predictor selection will affect the accuracy of the prediction. The parameters of the resulting model were further fine-tuned so that it was able to classify businesses from new real data. The research conducted showed that using the right type of graphical processing of input data, SMOTE technique and appropriate parameter settings, convolutional neural networks can predict the financial failure of companies with high accuracy.
|
|
|
Imbalanced data training approaches in neural network
Vicianová, Veronika ; Ředina, Richard (referee) ; Jakubíček, Roman (advisor)
This thesis deals with the research and implementation of methods that eliminate the influence of an imbalanced dataset on the learning of neural networks. Individual methods are compared with each other for different levels of imbalance. The experiments carried out in the work are also compared with the available literature and a control experiment, which was carried out without the method of eliminating the influence of an imbalanced dataset. The experiments are extended to another dataset containing the original imbalance and compared. In the theoretical section, the topic of neural networks and the problems that may occur during learning are brought up. Subsequently, convolutional networks and their optimization algorithms are presented. The thesis also contains a more detailed presentation of the issue of an imbalanced dataset, including the metrics used in experiments and approaches used to eliminate this problem.
|
guest ::
