Ensemble of Data-Driven Prognostic Algorithms with Weight Optimization and K-Fold Cross Validation
Ensemble of Data-Driven Prognostic Algorithms with Weight Optimization and K-Fold Cross Validation
复制标题
DOI：
10.1115/detc2010-29182
复制
发表时间：
2010-10
期刊：
影响因子：
--
通讯作者：
Chao Hu;B. Youn;Pingfeng Wang
中科院分区：
文献类型：
--
作者：
Chao Hu;B. Youn;Pingfeng Wang
传统的数据驱动预测方法是使用训练数据集构建多个候选算法，使用测试数据集评估它们各自的性能，并选择具有最佳性能的一个，而丢弃所有其他算法。该方法具有三个缺点：（i）所选择的独立算法可能不是鲁棒的，即，当在部署之后获取的真实的数据不同于测试数据时，其可能不太准确;（i i）其浪费了用于构造在部署中被丢弃的算法的资源;（iii）其除了训练数据之外还需要测试数据，这增加了算法选择的总费用。为了克服这些缺点，本文提出了一种集成数据驱动的预测方法，它结合了多个成员算法的加权和公式。提出了基于精度的加权、基于多样性的加权和基于优化的加权三种方法来确定数据驱动算法中成员算法的权重。采用k折交叉验证（CV）来估计加权方案所需的预测误差。两个案例研究，以证明所提出的预测方法的有效性。结果表明，集成方法与任何加权方案提供更准确的RUL预测相比，任何单一的算法和优化为基础的加权方案提供了最佳的整体性能之间的三个加权方案。
The traditional data-driven prognostic approach is to construct multiple candidate algorithms using a training data set, evaluate their respective performance using a testing data set, and select the one with the best performance while discarding all the others. This approach has three shortcomings: (i) the selected standalone algorithm may not be robust, i.e., it may be less accurate when the real data acquired after the deployment differs from the testing data; (ii) it wastes the resources for constructing the algorithms that are discarded in the deployment; (iii) it requires the testing data in addition to the training data, which increases the overall expenses for the algorithm selection. To overcome these drawbacks, this paper proposes an ensemble data-driven prognostic approach which combines multiple member algorithms with a weighted- sum formulation. Three weighting schemes, namely, the accuracy-based weighting, diversity-based weighting and optimization-based weighting, are proposed to determine the weights of member algorithms for data-driven prognostics. The k-fold cross validation (CV) is employed to estimate the prediction error required by the weighting schemes. Two case studies were employed to demonstrate the effectiveness of the proposed prognostic approach. The results suggest that the ensemble approach with any weighting scheme gives more accurate RUL predictions compared to any sole algorithm and that the optimization-based weighting scheme gives the best overall performance among the three weighting schemes.