저주파 통과 필터를 이용한 수온 예측과 K-평균 알고리즘을 활용한 수온 오측치 추정

초록

It is important to accurately predict it early and come up with a solution, since the increase in sea temperature has a direct relationship with global warming. Therefore, in this paper, we propose a method to predict water temperature more accurately with a prediction model using a low-pass filter in the frequency domain [6]. In general time-series prediction, Anomaly data ??are considered to be fatal factors that adversely affect learning. Therefore, in order to accurately predict the water temperature, we will also look at the method of detecting Anomaly data ??using autoregressive model and K-means algorithms. As the length of the data to be predicted through experiments increases, the model presented in this paper shows lower prediction error rate than the existing time series prediction model. And finally It was also confirmed that the performance of the model was higher when the data refined through Anomaly detection was applied to training.