A carregar...
Publicação
Variation of the relative stock prices positions on an abstract geometry space
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 669 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
The Markowitz theorem is the basis of current investment fund management. Based on time series pricing treatment, a portfolio is optimized by minimizing the risk for the same expected return or maximizing the return for the same risk. This approach assumes that time series have sections in Gaussian time, with a finite standard deviation and an expected stationary value, assumptions that are not verified. The approach proposed in this paper is to use unsupervised neural networks to make a distributed representation of stocks within a potential portfolio to build an abstract geometry space where the variations of the relative position of shares in that space will be studied. The work ends when the abstract space is built and shows the distance evolution of the set of actions in relation to the others.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Risk Analysis and Management
Palavras-chave
Portfolio management Modern Portfolio Theory Asset Correlation Artificial Neural Networks Machine Learning