Deep Learning Approaches for Predicting Stock Market Trends: A Comparative Study of New Zealand and Australian Markets

In the contemporary global economy, the complexity of financial activities has surged, making stock market predictions increasingly challenging due to inherent volatility and unpredictability. This paper presents a comparative analysis of various deep learning models—Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Networks (CNN), and hybrid models like CNN-LSTM and GRU-LSTM—for predicting stock market trends, focusing on the New Zealand and Australian markets. Traditional methods such as ARIMA and basic machine learning models have been surpassed by neural networks, which are adept at handling nonlinear financial data. While LSTM and GRU models excel in capturing long-term dependencies, they encounter challenges such as overfitting and slow training times. Hybrid models like CNN-LSTM and GRU-LSTM combine the strengths of CNNs and RNNs to enhance prediction accuracy. This study utilizes daily historical data from the S&P/NZX 50 and S&P/ASX 200 indices spanning from 2017 to 2023. Evaluation metrics—Root Mean Square Error (RMSE), R-squared (R2), and Mean Absolute Percentage Error (MAPE)—demonstrate that the GRU-LSTM hybrid model significantly outperforms others in predicting stock prices, achieving the lowest errors and highest accuracy. The findings underscore the potential of hybrid deep learning models in financial forecasting, providing valuable insights for investors and traders to make informed decisions. This research uniquely contributes to the field by being the first comprehensive comparative analysis of deep learning models specifically focused on the Australasian market region, addressing a significant gap in understanding how these models perform in different market conditions and scales.