Balanced portfolio via signed graphs and spectral clustering in the Brazilian stock market

A portfolio associated with a balanced signed graph that contains both positive and negative edges is more predictable and risk-averse, and is therefore likely to require less turnover. In this paper, we use a method based on spectral graph theory to determine whether a portfolio is balanced or not. Moreover, we combine this with spectral clustering to propose a strategy to build a balanced portfolio with hedging assets. This is applied to stocks listed on the Brazil Stock Exchange.