A Review on Deep Sequential Models for Forecasting Time Series Data

Table 5

A summary of published articles that used deep sequential models in weather predictions.

Ref.	Method	Application	Results obtained	Metrics used	Data unit
[76]	MINO-TCNN, MISO-TCNN	Weather forecasting	TCNN produced better forecasting. It can be applied as an effective method localized climate forecasting tool. It is executed on a stand-alone personal computer.	MSE	Every 15 minutes
[77]	ANN	Weather	The error in guessing 10 minutes in advance is the smallest statistically.	MSE, MAE, RMSE, and ME	10, 20, 30 minutes and 1 hour
[78]	ANN	Weather	System is capable of forecasting weather with a low error rate and a more acceptable structure	MSE ranges from 0.9325 to 3.5321	Hourly and daily
[79]	Hybrid ANN + PSO	Weather (wind speed)	The model was quite accurate.	MAPE (3–6%)	Daily
[80]	AEEMD-ANN	Rain fall	The model was shown to be effective in capturing very low SWM rainfall.	R, MAE, NRMSE, IA	Monthly
[81]	SSA-ARIMA-ANN	Rain fall	The hybrid model was capable of forecasting the catchment with a high degree of confidence.	R^2, RMSE, MAE, MPE, MNE	Daily
[82]	AGRU	Wind power forecasting (WPF)	The AGRU model proposed provides competitive capabilities in power system WPF. The attention method increases computing time every learning epoch, and hyperparameter tuning takes time	NRMSE and MAPE	Every 5 minutes