Predicting stock price movement using a stack of multi-sized filter maps and convolutional neural networks

Abstract

The paper explores the use of convolutional neural networks (CNNs) to predict the movement of the stock market from a classification perspective. Standard classification methods yield results with very low confidence and precision. We therefore propose a CNN enhanced by multi-size feature maps and spatial mapping that will provide a more accurate two-way classification for the collection of stocks. We also propose transforming stock indicators and data into a spatial map/image so that they can be processed using CNN. Our model and mapping fairs at an average of 80% weighted F1-score for a two-way classification of market movement. A trading strategy is also used and returns are compared to the benchmarks. The return of the proposed trading strategy for the period 2017 to 2020 is above the previous benchmarks. © 2022 Inderscience Enterprises Ltd.