A hybrid deep learning network for atmospheric carbon monoxide prediction in the Indian region
| Autor: | Sameer Poongadan, M. C. Lineesh |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
0209 industrial biotechnology
Series (mathematics) business.industry Deep learning Atmospheric carbon cycle Phase (waves) Monoxide 02 engineering and technology Hilbert–Huang transform 020901 industrial engineering & automation Singular value decomposition 0202 electrical engineering electronic engineering information engineering Environmental science 020201 artificial intelligence & image processing Artificial intelligence Time series business Algorithm |
| Zdroj: | INTERNATIONAL CONFERENCE ON COMPUTATIONAL SCIENCES-MODELLING, COMPUTING AND SOFT COMPUTING (CSMCS 2020). |
| ISSN: | 0094-243X |
| DOI: | 10.1063/5.0045784 |
| Popis: | Non-linear time series forecasting is of high significance since non-linearity occurs in most of the time series in practical situations. This study presents a hybrid forecasting procedure to predict the level of atmospheric Carbon monoxide in Indian region. The atmospheric Carbon monoxide prediction has a great significance since it’s increase can cause the deterioration of the atmosphere resulting in acute health issues for humans. Successful managing of Carbon monoxide demands an efficient model for predicting the future values of the Carbon monoxide. This paper presents a new time series forecasting EEMD-SVD- LSTM model which combines Ensemble Empirical Mode Decomposition (EEMD), Singular Value Decomposition (SVD) and Long Short Term Memory (LSTM) network to forecast Carbon monoxide data collected from Indian region. The model can be used for non-linear and non-stationary data. The model is a combination of three phases which are EEMD phase, SVD phase and LSTM phase. In first phase EEMD is applied to the data and decompose the data series into a finite number of IMF components and a residue. The second phase applies SVD to each IMF component to de-noise the same. In the third phase, LSTM is used to forecast each de-noised IMF component and residue. Then summing up all the forecast series we obtain the predicted values of the original series. The result of the proposed model is compared with other forecasting techniques such as LSTM, EMD-LSTM and EEMD-LSTM. The result shows that the proposed EEMD-SVD-LSTM model surpasses the other models in it’s efficiency to forecast the future values of carbon monoxide. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|