Abstract: Data mining for data with time attributes is called temporal data mining. The purpose of temporal data mining is to discover the knowledge of data in time. When the change of data is irregular, such as stock trading data, it is difficult to find valuable rules and laws. And neural network has the advantages of parallelism, fault tolerance, hardware implementation and self learning, this makes it a method for the application of stock classification forecasting. The stock data is converted into temporal stock data by combining stock data with temporal type. And then a classification method of temporal neural network is proposed. The stock data which is collected from some listed companies in ten years are analyzed, and the neural network classifier of temporal stock data is constructed. With the aid of this classifier，the stock data is classified and predicted. Experimental results show that the classifier has higher classification accuracy compared with the pre-improved neural network and support vector machine. The results also indicate that the neural network classifier with temporal data is very effective for the classification and prediction of multiple stocks, and it can be well applied to the classification prediction of the stock market.

Key words: temporal data, neural network;stock;classification forecasting

摘要： 对具有时间属性的数据进行数据挖掘称为时态数据挖掘，用以发现数据在时间上的知识，当数据变化不规律时，如股票交易数据,就很难发现有价值的规律与规则。而神经网络具有并行、容错、可以硬件实现以及自我学习的优点,可作为股票分类预测应用的一种方法。通过将股票数据与时态型相结合，将股票数据转换成时态型股票数据，提出时态神经网络模型的分类方法，对收集的若干上市公司十年内的股票数据进行分析，构建了时态股票数据神经网络分类器对股票进行分类预测。经过实验验证，相比改进前的神经网络和支持向量机方法，该分类器具有更高的分类准确率。结果证明，这种时态数据神经网络模型对于多只股票的分类预测是非常有效的，可以很好地运用到股票市场的分类预测中。

关键词: 时态数据, 神经网络, 股票, 分类预测