基于分段线性表示和高斯过程分类的股票转折点概率预测

针对股票交易过程中价格转折点的预测问题,提出了一种基于分段线性表示(PLR)与高斯过程分类(GPC)相结合的股票价格转折点预测算法PLR-GPC。该算法通过PLR提取股票历史价格序列的转折点,对转折点进行分类标记,建立基于GPC的股票价格转折点预测模型,以上述股票历史价格序列对模型进行训练,最终由预测模型对股票价格转折点进行预测,并对预测结果进行概率解释。将PLR-GPC与基于BP神经网络(BPN)的PLR-BPN算法、基于加权支持向量机支持向量机(WSVM)的PLR-WSVM算法进行实验对比:PLR-GPC在预测准确率上高于PLR-BPN与PLR-WSVM;在投资收益率上高于PLR-BPN,与PLR-WSVM持平。实验结果表明PLR-GPC在股票价格转折点的预测上是有效的,并且可以应用在实际股票投资交易中。

Integrating piecewise linear representation and Gaussian process classification for stock turning points prediction

Focusing on the prediction issue of the price turning point in stock trading process, a prediction algorithm of stock price turning point, named PLR-GPC, was proposed based on Piecewise Linear Representation (PLR) and Gaussian Process Classification (GPC). The algorithm extracted the turning points of the historical stock price series by PLR, and classified the points with different labels. A prediction model of the stock price turning point was built based on GPC, and it was trained with the turning points extracted by PLR. Eventually, the model could predict whether a new price would be a price turning point, and could explain the result with probability. An experiment on the real stock data was carried out among PLR-GPC, PLR-BPN (PLR-Back Propagation Network), and PLR-WSVM (PLR-Weighted Support Vector Machine). It showed that the PLR-GPC had higher forecast accuracy than the other two algorithms, and its rate of return was higher than PLR-BPN, almost equal to PLR-WSVM. The experimental result proves that the PLR-GPC is effective on stock turning point prediction and it can be applied in the actual stock investment trading.