Open AccessANALYSIS OF METHODS OF INCREASING DATA RELIABILITY FOR PROBLEMS OF SHORT TERM FORECASTING OF NODAL LOAD
Author(s) -
Pavlo Shymaniuk,
AUTHOR_ID,
V. Miroshnyk,
AUTHOR_ID
Publication year - 2021
Publication title -
pracì ìnstitutu elektrodinamìki nacìonalʹnoï akademìï nauk ukraïni/pracì ìnstitutu elektrodìnamiki nacìonalʹnoï akademìï nauk ukraïni
Language(s) - English
Resource type - Journals
eISSN - 1727-9895
pISSN - 2786-7064
DOI - 10.15407/publishing2021.60.051
Subject(s) - cluster analysis , reliability (semiconductor) , term (time) , data mining , series (stratigraphy) , computer science , dbscan , false positive paradox , time series , mathematics , algorithm , artificial intelligence , cure data clustering algorithm , correlation clustering , machine learning , paleontology , power (physics) , physics , quantum mechanics , biology
A comparative analysis of clustering methods was performed to identify gaps and anomalous values in the data. Data from the northwestern region of the United States were used for evaluation. According to the analysis results, it was found that the use of the DBSCAN method leads to a much smaller number of false positives. An algorithm for two-stage data validation using clustering and time series decomposition methods is proposed. Ref.9, fig. 3, tables 3.