PCE-BP: Polynomial Chaos Expansion-Based Bagging Prediction Model for the Data Modeling of Combine Harvesters

With the rapid developments of measurement and monitoring techniques, massive amounts of in-situ data have been recorded and collected from the measurement system of combine harvesters in their working process and/or field experiments. However, the relationship between the operation parameters and the performance index such as clearing loss usually changes greatly in different sample subspaces, which makes it difficult for conventional prediction models to model the in-situ data, since most of them assume that the relationship is the same or similar throughout the whole sample space. Therefore, a polynomial chaos expansion-based bagging prediction model (PCE-BP) is proposed in this article. A polynomial chaos expansion-based decision tree is constructed to divide the sample space such that the relationship between the operation parameters and the performance index in the same part is more similar than the others, and bagging is used to ensemble the polynomial chaos expansion-based decision trees to reduce the perturbation and provide robust predictions. The experiments on the mathematical functions show that the proposed prediction model outperforms polynomial chaos expansion, polynomial chaos expansion-based decision tree, and the conventional bagging prediction model. The proposed prediction model is validated through two monitoring datasets from a combine harvester. The experimental results show that the PCE-BP model provides better cleaning loss and impurity rate prediction results than the other prediction models in most experiments, showing the advantages of sample space partitioning and bagging in the data modeling of combine harvesters.

@article{Zhong2025,
  title     = {PCE-BP: Polynomial Chaos Expansion-Based Bagging Prediction Model for the Data Modeling of Combine Harvesters},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {Liangyi Zhong and Mengnan Deng and Maolin Shi and Ting Lou and Shaoyang Zhu and Jingwen Zhan and Zishang Li and Yi Ding},
  doi       = {10.14569/IJACSA.2025.0160190},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160190}
}