Financial Modelling System Using Deep Neural Networks (DNNs) for Financial Risk Assessments
Annotatsiya
ABSTRACT The FOREX market assessment is a big challenge for investors and global risk managers. However, the present study uses daily multicurrency exchange rate returns data from 2007 to 2022 to estimate the learning returns performance of the proposed model to find a safe‐haven currency for optimal investment strategy. The categorical returns are classified into good returns (GRs), bad returns (BRs) and no returns (NRs). Therefore, the present study needs to use a one‐hot‐encoding function to convert a categorical dataset into a numeric format with TensorFlow. The present study proposes a deep neural network‐based multilayer perceptron (DNN‐based MLP) with a backpropagation algorithm to estimate the learning returns performance of the proposed model to find a safe‐haven currency for optimal investment strategy. The findings showed that currency exchange rate return 2 (CERR2) is relatively a safe‐haven currency than currency exchange rate return 1 (CERR1) and currency exchange rate return 3 (CERR3). Moreover, the findings also showed that the proposed model gives optimal learning return performance. This study may assist FOREX investors to modify their investment strategies under shed light of findings of the study. In addition, the findings of the present study may also support global risk managers to revisit their hedging strategies.