摘要：There are many obvious evidences supporting a correlation between school facility and student behavior and performance. With the increasing awareness of sustainable school facility, incorporation of various operation cost impact into the consideration of school facility management is attracting a lot of attention. So the Green-School Project in South Korea aims to transform existing deteriorated elementary, middle, and high school facilities into eco-friendly energy saving schools through environmentally friendly materials and techniques and full-scale renovation and repair work. However, the total number of educational facilities in South Korea as of 2015 is 11,590 (5,978 elementary schools, 3,219 middle schools, and 2,393 high schools). Overall reconstruction of these deteriorated educational facilities is realistically difficult. Expenditure by school systems must stay within the limit of their available funding. So in order to plan exact operating cost, this paper presents a prediction improving method of the amount of electricity consumption of elementary school in South Korea by using two regressions, i.e., SVR (Support Vector Regression) and GPR (Gaussian Process Regression) and outlier detection methods, EE (Elliptic Envelope) and EM (Expectation and Maximization) algorithms. As a result, this study enables school facility managers to straightforwardly predict the electricity consumption of elementary school. This method can also extend to prediction of the amount of electricity usage for middle school and high school as well as elementary school.
关键词：Electricity Consumption; Elementary School; Operation Cost; Regression ;Algorithm; Support Vector Machine; Gaussian Process Regression; Elliptic Envelope; ;Expectation and Maximization
