Predicting company market values using artificial neural network (ANN) modelling

This master thesis focuses on the valuation of companies using a neural network. Its aim is to find an alternative valuation method that could complement classical methods. The final dataset was split into two parts in order to test the model on data that the model did not have the opportunity to observe. The target variable was chosen to be the capitalization of the company in May of the following year in order to build the model on data that corresponds to potential real use. At the end of the paper, the neural network is also compared with lasso regression, which has the function of a benchmark. The theoretical part of the thesis includes a literature review on company valuation and also describes the neural network architecture. The practical part describes the whole process of this project (data collection, data transformation and modelling). The results show that although the neural network achieves better results than lasso regression, its current performance is not yet sufficient. At the end of the paper, steps that could lead to better results are also suggested.