Power Consumption Analysis of an Educational Building using Rough Set Theory

• Main Authors: Brucal, Stanley Glenn E., Africa, Aaron Don M., De Jesus, Luigi Carlo M.
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; approximation; Buildings; Business; Consumer electronics; core reduction; Correlation; heuristic search k-fold cross; Humidity; lattice search; leaving-one-out; Power demand; ROSE2 software; Rough sets; Software
• ISSN: 2693-0854
• DOI: 10.1109/GCCE62371.2024.10760710

In this paper, rough set theory (RST) simulation was used to determine the correlation of school building power consumption with its pre-defined attributes or conditions, such as, monthly business operations parameters and outdoor thermal conditions. RST analysis has shown a higher classification accuracy when satisfactory description validation with combination of strength and similarity rule support is applied to both continuous and discretized datasets. Among all conditional attributes considered in the analysis, it was the number of school days and average outdoor relative humidity that have accurately predicted the building power consumption. When datasets are discretized, classification rate improved, but outdoor thermal conditions had minimal effect on the predictability of power consumption. The increase in dataset helped improve the classification rate but resulted in approximation and core reduction results.