A Mean Weighted Squared Error-based Neural Classifier for Intelligent Pattern Recognition in Smart Grids
Supervised learning is widely used in pattern recognition and classification due to its strong ability to enhance data accuracy. Loss functions have proven to be a critical factor in enhancing the predictive accuracy of intelligent classifiers with diverse architectures and characteristics. This pap...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | International Journal of Electrical Power & Energy Systems |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S0142061525005204 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Supervised learning is widely used in pattern recognition and classification due to its strong ability to enhance data accuracy. Loss functions have proven to be a critical factor in enhancing the predictive accuracy of intelligent classifiers with diverse architectures and characteristics. This paper introduces a new extension of the conventional Mean Squared Error loss function, called Mean Weighted Squared Error (MWSE), specifically designed for renewable energy classification purposes. In the proposed MWSE, unequal weights are assigned to each estimated data point, unlike the conventional version. In this paper, Multilayer Perceptron Neural Networks (MLPs) are employed to implement the proposed MWSE loss function. To assess the effectiveness of the proposed MWSE-MLP classifier, a total of four benchmark datasets related to the renewable energy have been utilized. The empirical findings demonstrate that the proposed classifier consistently outperforms conventional shallow/deep intelligent classifier across all case studies. On average, the proposed MWSE-MLP classifier achieved an impressive classification rate of 96.21%, which is 1.98% higher than that of the classic MLPs. In addition, numerical results of an extensively reviewed case study indicate that the proposed classifier can also yield more accurate results than some other well-known shallow intelligent classifiers such as support vector machine, random forest, and decision tree by 2.72%, 3.34%, and 3.82% improvement, respectively. These improvements are not limited to shallow intelligent classifiers, but also repeated for deep learning classifiers, such as the long short-term memory, transformers and convolutional deep neural networks. |
|---|---|
| ISSN: | 0142-0615 |