预测模型可靠性的模糊数学评价方法

预测模型的可靠程度是通过预测结果中分布规律的可信度体现出来的。针对常见的预测模型可靠性评价中存在的问题,将预测模型预测结果的可信概率定义为预测模型的可靠度,提出了一种评价预测模型的新方法。在新方法中,运用模糊数学理论对预测结果的可信程度进行了评价,建立了预测结果可信度与预测结果相对误差绝对值之间的隶属函数关系,并将模糊数学与可靠性理论相结合,给出了求解预测模型可靠度的计算公式。以含蜡原油粘温关系模型为例,对新方法的评价过程进行了验证。结果表明,对同一种油样采用不同的隶属函数,或对不同油样采用同一个隶属函数,所得预测模型的可靠度均不相同,这说明该方法具有通用性。

A new assessment method for reliability of prediction model with fuzzy mathematics

The distribution of the authentic forecast results can embody the fiduciary level of the prediction model. The probability of the authentic forecast results obtained by the prediction model was defined as the fiduciary level of prediction model. A new method for assessment of the fiduciary level of prediction model was proposed. In order to assess the fiduciary level of the forecast results, a membership function for describing the relationship between the fiduciary level and absolute value of relative error of forecast results was established on the theory of fuzzy mathematics. By using the fuzzy mathematics and reliability theory, the formula to calculate the fiduciary level of the prediction model was provided. A prediction model for waxy oil viscosity was taken as an example to prove the applicability of the assessment method. The results show that the fiduciary levels of prediction model are different for the same oil sample with the different membership function or for the different oil sample with the same membership function.