Yapay Zeka Modeliyle Genişletilmiş Hibrit Black-Litterman Model Önerisi, Borsa İstanbul Bıst-30 Endeks Verileri İle Test Edilmesi

Abstract

In this study a new hybrid model developed using both artificial intelligence algorithms and GARCH econometric model. This hybrid model is used to produce daily basis investor views these are rational, systematic and based on mathematical calculations, for Black-Litterman portfolio optimization model. These investor views used at Black-Litterman model and applied to common stocks that are in BIST 30 index of Borsa Istanbul market to test the validity and reliability of the hybrid model. The hybrid model is composed of two stages. At the first stage, technical indicators of common stocks were forecasted on a rolling daily basis using its historical data. At the second stage, a support vector machine model was built based on technical indicators as input and common stocks’ returns as output. The technical indicator forecasts made by GARCH model were used as input at artificial intelligence model and common stocks’ returns forecast were produced. These forecasts were used as absolute investor views at Black-Litterman optimization model. As a result, higher return values achieved by this hybrid model compared to BIST 30 index returns, equally weighted portfolio returns or random weighted simulation based portfolio returns. In addition, it is found that the best holding period for the investors is 20 business days according to the calculations made through hybrid model results.