Using Convolutional Neural Networks for Predicting Stock Market Returns: A Comparative Analysis

Abstract

Predicting stock market returns is one of the fundamental challenges in finance and investment. Convolutional Neural Networks (CNNs), due to their high ability to identify complex patterns and nonlinear relationships, have gained attention as an efficient method for predicting financial data. With their multilayered structure, CNNs can analyze large and multidimensional datasets, outperforming traditional models in forecasting stock market fluctuations. This study aims to evaluate and compare the performance of Convolutional Neural Networks (CNNs) with other artificial intelligence models and traditional methods in predicting stock market returns. This study employs a literature review and descriptive analysis approach. Financial data, including stock prices, trading volumes, and economic indicators, were analyzed, and the results of CNNs were compared with other predictive models such as linear regression, Support Vector Machines (SVM), and ARIMA models. The results indicate that CNNs provide higher accuracy in predicting stock market returns compared to traditional methods, effectively identifying hidden and complex patterns in the data. CNNs, due to their multilayered structure, offer more precise results in unstable and volatile environments. However, challenges such as the need for large datasets and substantial computational resources were also noted in using this method. CNNs, as a powerful deep learning model, can improve the accuracy of financial forecasts, but optimization is needed in scenarios with limited data and computational resources.