Scholarlink

A problem of an optimal combining of different computational intelligence models’ outputs is considered for the multivariate case. An ensemble approach is used as a solution method. Specifically, a new bagging procedure is proposed that provides an optimal solution both in offline (batch) and online (sequential) modes. The latter is also applicable to non-stationary data processing and has a high learning rate. The proposed method is tested on the multivariate short-term electric load forecasting problem with an ensemble containing 3 models with 4 outputs. Experimental results support theoretical findings. The introduced system can be used to process data of various nature using different types of models (including shallow and deep networks, as well as hybrid systems) as ensemble members. There is a possibility to calculate fuzzy membership degrees of each ensemble member’s output to the optimal result, which can be used to track changes in the multivariate signal under consideration.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive Bagging of Hybrid Systems of Computational Intelligence Using Metamodel Online Optimal Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yevgeniy Bodyanskiy,
  • Olha Chala,
  • Iryna Pliss,
  • Sergiy Popov

摘要

A problem of an optimal combining of different computational intelligence models’ outputs is considered for the multivariate case. An ensemble approach is used as a solution method. Specifically, a new bagging procedure is proposed that provides an optimal solution both in offline (batch) and online (sequential) modes. The latter is also applicable to non-stationary data processing and has a high learning rate. The proposed method is tested on the multivariate short-term electric load forecasting problem with an ensemble containing 3 models with 4 outputs. Experimental results support theoretical findings. The introduced system can be used to process data of various nature using different types of models (including shallow and deep networks, as well as hybrid systems) as ensemble members. There is a possibility to calculate fuzzy membership degrees of each ensemble member’s output to the optimal result, which can be used to track changes in the multivariate signal under consideration.