基于信号分解和机器学习模型的股票价格预测

Stock Price Prediction Based on Signal Decomposition and Machine Learning Models

摘要: 随着全球经济的发展，人们对投资不同领域的标的资产表现出了浓厚兴趣，其中，投资股市是最受所有投资者欢迎的方式之一。为了获得额外的收益，投资者尝试使用不同的数学和统计方法来预测他们关注的股票的趋势或价格。在这项研究中，为了提高股票预测的效果，我们首先使用不同的信号分解方法，如经验模式分解(EMD)、集合经验模式分解(EEMD)和完全自适应噪声集合经验模态分解(CEEMDAN)来减少原始股票价格数据的噪声部分。以信号数据作为机器学习模型的输入，我们能够根据现有金融市场的股票交易数据，通过滑动窗口法对过去几年的股票价格进行预测。最后的结果表明，模型在预测不同信号分解的时间序列时有良好的表现。

Abstract: With the development of the global economy, people have more interest in investing different fields of underlying assets, among which the stock market is one of the most popular ways for all the investors. In order to gain extra benefits, investors try to use different mathematics and statistics methods to predict the tendency or prices of the stock they prefer. In this research, in order to improve the effects of stock prediction, first we use different methods of signal decomposition such as Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD) and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to reduce the noise of original stock price. With the signal data as input of machine learning models, we are able to forecast the stock price in the last few years by sliding window method based on stock trading data in existing financial markets. The last result shows that models have good performance in forecasting time series with different signal decomposition.