Abstract: Economic early warning helps decision-making by judging the tendency of economic development. In practical early warning, it is not convenient to deal with early warning of multi-value, and some economic data of indicators are absent. Both of them are the main difficulties that the economic early warning is facing with. This paper introduces maximum entropy in economic early warning. Among all the consistent predictions, it regards the probability is predominant when the entropy is maximum. It uses the strategy of feature correction and probability to solve the above problems. Meanwhile, it overcomes the independence hypothesis and has a high speed. Experiments testified the validity of the method.

Key words: Economic Early Warning, Maximum Entropy, Pattern Recognition, Feature Correction

摘要： 经济预警通过对经济发展趋势进行综合的判断，可以辅助人工决策。然而在实际经济预警中，经常出现多值警度难于方便处理以及某些经济指标的数据无法搜集完全的问题，这也是经济预警当前面临的主要问题。本文将最大熵方法引入经济预警，认为在对预警警度的所有相容的预测中，熵最大的预测出现的概率占绝对优势，以此通过特征补偿的方法有效处理了经济预警中经济数据缺失的问题，并通过计算各个警度出现的概率解决多值经济警度问题；同时，此方法无须满足特征独立性假设，训练速度快。最后，通过实验验证了该方法的有效性和高效性

关键词: 经济预警, 最大熵, 模式识别, 特征补偿