Predicting Stock Market Index and Credit Default Swap Spreads using Artificial Intelligence and Determining Non-Linear Relations

Abstract

In this study, a simulation model has been established to forecast the stock price index of Borsa Istanbul (BIST100) and 5 year maturity credit default swap spreads (CDSs) with an artificial intelligence approach. In the study where short-term and long-term relationships between variables were examined using non-linear econometric models such as Kapetanios, Shin, and Snell (KSS) and Exponential Smooth Transition Autoregressive (ESTAR) Vector Error Correction Model, Türkiye's 1211 data set obtained were used in the period 08.21.2015 - 08.17.2020. With this data set, a multilayer perceptron artificial neural network model has been established. Levenberg-Marquardt algorithm was used as the training algorithm in the feedforward backpropagation network model with 25 neurons in the hidden layer. Six input variables, which are considered to be common parameters affecting the target values, were defined in the input layer and BIST100 and CDSs were predicted at the output layer. The performance analysis of the network model was completed using various performance parameters, and in addition, a comprehensive performance analysis was performed by comparing the simulation results gotten from the neural network with the target values. The model was able to predict BIST100 with an average deviation of 0.04% and CDSs with an average error of -0.163%. These error rates indicate that the developed artificial neural network has been designed to predict BIST100 and CDSs with ideal accuracy.