【摘要】 针对企业、事业单位的财务金融系统存在非线性、非平稳性与噪声大的特点,为提高财务金融长期趋势的预测准确率与可靠性,提出了一种基于粗糙集的财务金融时间序列预测算法。通过粗糙集的化简概念删除输入数据集的冗余信息,根据事实信息生成无噪声的决策规则。考虑到训练样本的时期远近对粗糙集模型的分类准确率存在影响,设计了加权调和的粗糙集模型,为时期久远的训练样本分配较低的权重,为时期较近的训练样本分配较高的权重,提高近期训练样本对粗糙集模型的贡献。此外,提出了基于时间加权调和的决策冲突方案。基于香港恒生指数的实验结果表明:相比原粗糙集模型与支持向量机模型,本算法获得了更高的预测准确率。更多还原

【Abstract】 The financial systems of companies and public service institutions are non-linear,nonstationary and noisy. In order to improve the accuracy and the reliability of prediction of long term trend of financial time sequence,a financial time sequence prediction algorithm based on the rough set is proposed. The concept of deduct of rough set is adopted to remove redundant information from the input data set,the decision rules with no noise are generated according to the factual information.Because the dates of the training samples have an influence on the accuracy of the rough set,a weighted rough set model is designed,the old training samples are assigned small weight values and the near training samples are assigned big weight values,so that the recent training samples contribute more to the rough set model. Besides,a decision conflict schema based on weighted summation of time is proposed. The experimental results based on the HK hang seng index show that the proposed algorithm performs better prediction accuracy than the original rough set model and the original support vector machine.更多还原