An Economic Forecasting Method Based on the LightGBM-Optimized LSTM and Time-Series Model.

Area of Science:

• Computational Finance
• Machine Learning
• Time Series Analysis

Background:

• Accurate stock price prediction is crucial for financial decision-making and market supervision.
• Stock price fluctuations are influenced by complex, dynamic factors, exhibiting inherent randomness.
• Traditional time-series and machine learning models have limitations in capturing stock market dynamics.

Purpose of the Study:

• To investigate the efficacy of Long Short-Term Memory (LSTM) algorithms for stock market forecasting.
• To develop an optimized LSTM model by integrating LightGBM for improved short-term stock price prediction.
• To compare the performance of the proposed LightGBM-LSTM model against Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) models.

Main Methods:

• Utilized the Long Short-Term Memory (LSTM) algorithm for stock closing price prediction.
• Developed a novel LightGBM-optimized LSTM model to address shortcomings of the general LSTM.
• Empirically evaluated the LightGBM-LSTM, RNN, and GRU models using Shanghai and Shenzhen 300 index data.

Main Results:

• The LightGBM-LSTM model demonstrated superior prediction accuracy compared to RNN and GRU.
• The proposed model exhibited a greater ability to track stock index price trends effectively.
• Experimental results confirmed the enhanced performance of the LightGBM-LSTM over GRU and RNN algorithms.

Conclusions:

• The LightGBM-LSTM model offers a significant advancement in short-term stock price forecasting.
• This optimized approach provides more reliable investment decision-making information.
• LSTM-based models, particularly when optimized, are highly suitable for complex stock market analysis.