Özet:

Investors aim to strengthen their portfolios and achieve higher returns by allocating their savings to assets with high return potential. Among various financial instruments, stocks are considered investment vehicles that offer a high potential for returns. Consequently, stock price prediction has been a subject of curiosity for many investors and researchers and continues to be so. However, the pursuit of high returns also introduces high risks and uncertainties. The fundamental sources of this uncertainty include political events, social developments, international relations, and sectoral changes, all of which influence stock prices. The prediction of stock price movements, which is inherently challenging, becomes even more difficult due to the influence of versatile and complex factors. Traditionally, stock price predictions have been based on statistical and econometric methods. In recent years, advancements in technology and computer sciences have led to the increasing use of artificial intelligence and machine learning methodologies in stock price predictions. Since the development of deep learning methods in the 2000s, complex relationships in stock price movements have been identified, yielding successful results. This study aims to predict stock prices using multivariate Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN) models, and to compare the performance of these models. To achieve this goal, analyses were conducted using the daily closing stock prices series of Vestel Elektronik Sanayi ve Ticaret A.Ş. (VESTL) between June 3, 2013, and May 28, 2021, along with independent variables that could affect this series. According to the analysis results, comparing the mean squared error (MSE) and the square root of the mean squared error (RMSE) metrics, it was determined that the CNN model performed the best, and the RMSE value for this model was calculated as 0.086.

Anahtar Kelimeler: