We apply concepts form machine learning to forecast stock prices. First, we introduce the general (3 by 3) forecasting model, in which the financial markets are populated by three types of stocks: Overpriced stocks, underpriced stocks and fairly priced stocks. The objective of the financial analyst or a potential investor is to identify which stock belongs to which classification, and to take the relevant investment decision. Second, we present two different numerical examples to illustrate our forecasting models, and estimate all the relevant statistics, as well as the forecasting accuracy. Third, we introduce the Receiver Operator Curve (ROC), and demonstrate the trade-off between the sensitivity and specificity of the prediction. We also discuss the performance evaluation of forecasting stock prices.

我们采用机器学习的概念来预测股票价格。首先，我们介绍一般的（3 x 3）预测模型，其中金融市场由三种类型的股票组成：定价过高的股票，定价偏低的股票和定价合理的股票。财务分析师或潜在投资者的目的是确定哪些股票属于哪种分类，并做出相关的投资决策。其次，我们提供两个不同的数值示例来说明我们的预测模型，并估计所有相关统计数据以及预测准确性。第三，我们介绍了接收方操作员曲线（ROC），并演示了预测的敏感性和特异性之间的权衡。我们还将讨论预测股价的绩效评估。