This paper builds two synthetic models according to single model, one is combination optimal fitting model in order to reduce the fitting residual error as the goal, the other is combination optimal prediction model in order to increase forecasting ability. The optimal fitting model uses two⁃times programming algorithm to as⁃sign the weights of the single model base and the minimum varince of historical data. Based on the optimal fitting model, the weight pro⁃gram in order to increase forecasting ability is set to get prediction model. The model is maximum close to the law of historical data change, and makes up the deficiency of the fitting model in the development trend. Error analysis method is used to compare the advantages and disadvantages between the models. At last, through the actual example, the effectiveness of synthetic optimal forecasting model is compared and confirmed.%根据单一模型建立2种综合模型，分别为以减少拟合残差为目标的综合最优拟合模型和以提升预测能力为目标的综合最优预测模型。综合最优拟合模型采用二次规划算法对单一模型库进行与历史数据方差最小为目标的权值分配，然后在综合最优拟合模型的基础上以提升预测能力为目标的再次权值规划得到综合最优预测模型。通过综合最优拟合模型建立的综合最优预测模型不但在最大程度上接近历史变化规律，并弥补拟合模型在发展趋势上的不足。综合最优预测模型使用误差分析方法对文中所用模型进行分析、比较以区别模型优劣性，最后应用实际的算例，比较确定综合最优预测模型的有效性。
展开▼
