CTGAN-based oversampling and cost-sensitive deep neural network to predict smart money activity in stock market

Investment in the stock market has become a trend in today’s era. The primary force moving the market in a specific direction is the large buying and selling of hedge funds, pension funds, banks, etc. This paper proposes a deep learning strategy to predict smart money activity. Initially, a framework composed of technical analysis and quantitative analysis is considered to create a dataset. Then using this framework a comprehensive dataset is built from the data available in the National Stock Exchange of India (https://www.nseindia.com/). In the proposed approach, minority samples in the dataset are first oversampled using Conditional Tabular Generative Adversarial Network based approach to even out class imbalance in the real-life dataset. Using the modified dataset, a cost-sensitive deep neural network is trained to predict smart money activity in the stock market. The proposed method is assessed using different evaluation metrics and the findings validate the superiority of the proposed methodology.