Stock market prediction by combining CNNs trained on multiple time frames

This paper explores a different method used for market analysis in the Forex stock market. Various econometric models, moving averages, technical indicators, and machine learning techniques have been investigated for predicting stock market trends. This study focuses on designing a new model called the multi-CNN model, which incorporates domain knowledge of Forex. The model is evaluated using EURUSD data from January 2015 to December 2020. The data is preprocessed, normalized, and divided into training, validation, and testing sets. The performance of the proposed model is compared with benchmark models such as Single-LSTM, Single-GRU, and Single-CNN. The results indicate the promising performance of the multi-CNN model in stock market forecasting. The paper provides insights into applying deep learning approaches for predicting stock market trends, highlighting the advantages of combining CNNs and utilizing multiple time frames over simple models such as simple CNN, LSTM, and other recurrent neural network-based models.