http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
PM<SUB>10</SUB> 농도 예측을 위한 머신러닝 기반 결측치 처리의 실증적 분석
이주현(Juhyun Lee),이윤관(Younkwan Lee),홍유진(Yoojin Hong),전문구(Moongu Jeon) 대한전자공학회 2019 대한전자공학회 학술대회 Vol.2019 No.6
The data including meteorology and air pollutants data for forecasting PM10 concentration recorded by monitoring stations contained various number of missing values due to measurement device defects or inability to measure by natural disasters. Therefore, it is very important to reasonably fill these missing values to improve the accuracy of PM10 concentration prediction. In this paper, we discuss a variety of machine learning based methods including Linear and Non-linear techniques to handle missing data. We use 5 methods to deal with missing values, and for each case creates datasets containing 10 types of weather information collected in Seoul area for predicting PM10, and we compare the PM10 concentration prediction performance using the datasets based on the Long Short-Term Memory Neural Network. Experiments show that the Non-linear imputation methods achieved significantly improved performance in PM10 concentration prediction compare to the linear imputation methods.