Using Multivariate Time Series Data via Long-Short Term Memory Network for Temperature Forecasting
Abstract:
This paper presents a Long-Short Term Memory, as a special kind of Recurrent Neural Network, capable of learning long-term dependencies to estimate the temperature. In order to improve the performance of the proposed model, a multivariate time series is considered. A Multivariate time series has more than one time-dependent variable. Each variable depends not only on its past values but also has some dependency on other variables (i.e., temperature, humidity, atmosphere pressure, wind speed, cloud cover and among others). A pre-processing is applied to dataset provided by the National Institute of Meteorology and Hydrology to obtain daily weather variables. The proposed model is compared with a previous approach using standard metrics, obtaining better results on the average temperature in Guayaquil city.
Año de publicación:
2021
Keywords:
- TEMPERATURE
- long short-Term memory
- pbkp_rediction
- Recurrent neural networks
- Multivariate Time Series
Fuente:


Tipo de documento:
Conference Object
Estado:
Acceso restringido
Áreas de conocimiento:
- Aprendizaje automático
- Ciencias de la computación
- Meteorología
Áreas temáticas:
- Ciencias de la computación