We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Statistical Features Extraction and Performance Analysis of Supervised Classifiers for Non-Intrusive Load Monitoring.
- Authors
Hasan, Md. Mehedi; Chowdhury, Dhiman; Khalid Hasan, Abu Shahir Md.
- Abstract
An approach to extract distinctive statistical features embedded in current and power signatures of different electrical appliances to substantiate efficacious classification for non-intrusive load monitoring (NILM) is presented in this letter. Supervised classifiers - naıve Bayes, multi-class support vector machine (SVM), ensemble, binary decision tree (DT) and discriminant analysis are employed for performance evaluation based on the extracted feature values. The testbed is COOLL NILM public dataset constituted by 42 devices of different power ratings. The training and testing accuracies along with cross-validation losses associated with each classification algorithm are determined. As a comparative analysis, binary DT classifier produces the best results. Performance assessment corroborates the reliability of the proposed framework for NILM applications.
- Subjects
SUPPORT vector machines; CLASSIFICATION algorithms; DISCRIMINANT analysis; DECISION trees; FEATURE extraction; PERFORMANCE evaluation
- Publication
Engineering Letters, 2019, Vol 27, Issue 4, p1
- ISSN
1816-093X
- Publication type
Article