Estimation of risk and stock returns using a combined approach of planning logarithmic fuzzy preferences and neural networks

The present research aims to identify the influential variables on stock portfolio selection, prioritize these variables, and estimate the risk and return of sample stocks using neural network algorithms.

Rationale: 

Stock portfolio selection has always been an intriguing and practical issue in financial matters and financial markets. In order to address the existing drawbacks in research related to stock portfolio selection, the idea of employing the fuzzy logarithmic preference programming method for analyzing factors affecting stock portfolio selection and utilizing neural networks for risk and return estimation is reinforced.

The present research offers a novel combined approach for stock portfolio selection consisting of two stages: In the first stage, by conducting interviews with experts and examining available documents and records, six primary criteria for selecting an optimal stock portfolio are identified. Using the fuzzy logarithmic preference programming approach, the weights of these criteria are determined. In the second stage, the risk and return of stocks are predicted using neural network algorithms.

The findings indicate that profitability, efficiency, and risk are the most important criteria in selecting an optimal stock portfolio, respectively. Additionally, the designed neural network successfully fitted the returns and risks of stocks.