Stock picking with machine learning

• Main Author: Zhu, Bangyuan
• Other Authors: Wang Lipo
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Stock; Deep learning; Artificial intelligence
• Online Access: https://hdl.handle.net/10356/176936
• Institution: Nanyang Technological University

In recent years, deep learning models based on neural networks have revolutionized the field of artificial intelligence. Unlike traditional machine learning approaches, deep learning models extract intricate features from raw data and perform end-to-end predictions, minimizing manual intervention and information loss associated with multi-step learning processes. However, when applied to multi-feature stock selection tasks, existing deep learning models may encounter limitations, necessitating the design of customized network structures for optimal effectiveness. This research proposes a novel approach inspired by convolutional neural networks to address the challenges of stock selection. A feature extraction layer akin to a convolutional layer is constructed by developing custom operator functions. When integrated with batch normalization, pooling, and fully connected layers, it extracts features from volumetric and pricing data for return rate prediction. The experimental results demonstrate the efficacy of the proposed model in enhancing stock selection accuracy and profitability metrics compared to conventional deep learning architectures. The study showcases the potential of customized network structures in improving predictive performance and decision-making processes in investment strategies.