A Neural Network-based Stock Timing Model

The stock market is inherently complex and volatile, making stock prediction crucial for investors. Traditional forecasting methods often struggle to adapt to the ever-changing nature of market dynamics. This paper proposes a neural network model for stock timing, aiming to predict stock price trends and provide actionable insights for investment strategies. We utilize historical daily trading data of Sinopec (600028.SH) from 20010808 to 20240618, including key features such as opening price, closing price, highest price, lowest price, and trading volume. The data undergoes thorough preprocessing, including missing value imputation, outlier detection, and normalization, ensuring that the neural network model can learn effectively from the time-series data. The model is built using a Multi-Layer Perceptron (MLP) architecture, with the Sigmoid activation function for binary classification of stock price movement trends. The model’s performance is evaluated based on accuracy, with results indicating a 37% error rate. This study demonstrates the potential of neural networks to enhance stock prediction accuracy while emphasizing the importance of continuous model refine-ment and the incorporation of up-to-date market data for improved forecasting.