以中信海直为例的不同时间维度股票预测研究

Research on Stock Forecasting in Different Time Dimensions with CITIC Haizhi as an Example

摘要: 本研究旨在评估多种机器学习、深度学习和统计分析技术在股票价格预测领域的效果。通过选择中信海直(股票代码：000099)的1分钟、15分钟和日线数据，应用ARIMA、LSTM、RNN和CNN等模型进行深入分析。我们通过计算均方根误差(RMSE)、平均绝对误差(MAE)和确定系数(R²)等指标来评估模型性能，发现使用高时间密度数据(如1分钟数据)的预测精度明显优于使用低时间密度数据(如日线数据)。具体表现在较低的MAE值和偏差，这说明模型在处理更频繁的数据时能够更准确地捕捉股价变动。本研究结果不仅证实了在股价预测中数据的时间密度是一个关键因素，也为投资者提供了更准确的市场趋势预测，帮助他们做出更明智的投资决策，从而可能最大化投资回报。

Abstract: This study aims to evaluate the effectiveness of various machine learning, deep learning, and statistical analysis techniques in the field of stock price prediction. By selecting one-minute, fifteen-minute, and daily data from CITIC Haizhi (stock code: 000099), we conducted an in-depth analysis using models such as ARIMA, LSTM, RNN, and CNN. We assessed the performance of these models by calculating metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results reveal that the predictive accuracy using high-frequency data (such as one-minute intervals) significantly surpasses that using low-frequency data (such as daily intervals). This is evident in the lower MAE values and smaller deviations, indicating that the models can more accurately capture stock price movements with more frequent data inputs. The findings not only confirm that data frequency is a critical factor in stock price prediction but also provide investors with more accurate market trend forecasts, aiding them in making smarter investment decisions and potentially maximizing investment returns.