Applying Convolutional Neural Networks to Stock Market Forecasting – A Case Study of Stock Volume Prediction

Stock market forecasting is always a challenge because of dynamic, non-linear, non-parametric and high-noise features of stock data. Researchers have been applying many statistical methods to handle complex financial data, but since DL techniques have emerged, Artificial neural networks have been successfully utilized for automatic feature extraction and prediction. Among many ANN tools, convolutional neural networks (CNN) prove to produce better learning capabilities and therefore manage big data in capital markets, enabling stock market prediction. Therefore, with this work authors seek to explore the possibility of using CNN to process graphics containing stock charts and demonstrate the potential of predicting volume changes. The aim of the study is to develop a CNN’s based model that supports prediction of stock volume changes in the capital markets. Moreover, the authors intend to compare the quality measures of the generated predictive model for different types of stock charts. In relation to the aforementioned, a noteworthy contribution to the existing body of knowledge involves comparing the utility of two types of charts for analyzing volume values.