基于组合Boosting回归的软测量建模

Soft sensing modeling based on combined Boosting algorithm

为了提高Boosting回归算法的稳定性,提出了动态加权的组合Boosting回归算法,即DA-Boosting算法。首先以BP神经网络作为弱学习器,再调用Boosting回归算法构造强学习器,最后以强学习器得到的回归函数作为子函数进行动态加权平均,得到最终的组合函数。几个经典的分析回归数据集的测试表明,该算法不但具有良好的泛化能力,而且泛化性能稳定。最后将DA-Boosting算法用于丙烯软测量建模,应用结果表明该软测量模型泛化性能好,测量精度高。

To improve the Boosting algorithm stability, a Dynamically Averaging Boosting algorithm（DA-Boosting） is proposed.Using BP neural networks as a base learner, a leveraging learner is constructed by Boosting algorithm.The DA-Boosting algorithm is obtained by dynamically averaging the regression functions trained by leveraging learners.Experimental studies on three typical regression datasets show that this algorithm has good and stable generalization ability.Finally the DA-Boosting algorithm is applied to construct a soft sensing model for propylene concentration.Application results show that this model has high measurement precision as well as generalization ability.