WEKO3
アイテム
k-Nearest Neighbors for Univariate Time Series Prediction with Missing Values
http://hdl.handle.net/10458/00010053
http://hdl.handle.net/10458/00010053d70c2f15-061d-4bb7-bbcf-3b38d5ace8fc
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
k-Nearest Neighbors for Univariate Time Series Prediction with Missing Values (571.5 kB)
|
|
| Item type | その他 / Others(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2020-10-28 | |||||
| タイトル | ||||||
| タイトル | k-Nearest Neighbors for Univariate Time Series Prediction with Missing Values | |||||
| 言語 | en | |||||
| 言語 | ||||||
| 言語 | eng | |||||
| 資源タイプ | ||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_1843 | |||||
| 資源タイプ | other | |||||
| 著者 |
Hongsung, S.
× Hongsung, S.× Chawalitsuwan, C.× Chujai, P. |
|||||
| 抄録 | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | The objective of this research is to find the optimal model for univariate time series data with the missing values that occurred at the collection step. At this stage, we fixed the missing values problem by applying the CART algorithm in the MICE package of R language. After imputing the missing values, we constructed the daily time series model with six algorithms, namely k-Nearest Neighbors （kNN）, Naive forecasting method, Holt's trend method, Automated exponential smoothing forecasts, Simple exponential smoothing, and Autoregressive Integrated Moving Average Model （ARIMA）. The data used in this study is an hourly time series level of carbon monoxide （CO） data, which is an air pollutant that was collected during the period of 01/01/2012 to 31/08/2018. Selection of suitable models is determined by the performance of forecasting with RMSE and MAPE values. The results showed that the optimal model for prediction the daily univariate time series is the kNN model. This model was able to predict the best daily time series data three days in advance. The secondary predictive models were ARIMA （6,1,6）, Holt's trend method, Simple exponential smoothing, Naive forecasting method, and Simple exponential smoothing, respectively. The results of such trials, the concept of finding a suitable model for univariate time series data with the missing value can be applied to other data sets with such characteristics. | |||||
| 言語 | en | |||||
| 書誌情報 |
en : International Conference on Science, Technology and Education 2020 (ICSTE 2020) p. 100-104, 発行日 2020-09 |
|||||
| 著者版フラグ | ||||||
| 出版タイプ | VoR | |||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||
