Stock market prediction with optimized PLSTM-AL in smart urban cities

Abstract

Accurate stock market forecasting is crucial for investors and participants, as even small improvements in prediction accuracy can significantly impact revenue. However, forecasting is challenging due to the noise, complexity, and volatility inherent in stock data. Recent advancements in deep learning have led to the development of robust models for sequence prediction. This study introduces PLSTM-AL, a novel stock market forecasting model that integrates Pareto-like sequential sampling optimization, Long Short-Term Memory (LSTM), and an Attention Layer. This combination enhances forecasting accuracy, particularly for stock market direction. The model is tested using daily data from five global indices spanning 2003-2024, and the prediction results were compared with other sophisticated deep learning models like Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) models. Results demonstrate that PLSTM-AL outperforms these models, achieving high evaluation scores across most datasets, hence, confirming its robustness and effectiveness in stock market prediction.