Using Principal Components Analysis and Neural Networks for Improving the Prediction of Weather Parameters

Noura Alhussein(1)  , M. Taher Anan (2) and Yahia Mohamad Fareed (1)

  • of Statistics, Faculty of Science, University of Aleppo.Aleppo-Syria
  • of Communication Engineering, Faculty of Electrical and Electronic Engineering, University of Aleppo. Aleppo-Syria

(*Corresponding author:  00963934050014،

 Received: 05/01/2020                   Accepted: 06/02/2020


In this paper, Principal Components Analysis approach (PCA) and the Multilayer Feed Forward Neural Networks (MLFF) have been hybridized to improve the prediction of monthly rates of minimum, maximum and dry temperatures, wind speed, air pressure, relative humidity, and the monthly total rainfall of Aleppo city in Syria, and the  monthly rates of minimum, maximum and dry temperatures, vapor pressure, and rain precipitation of the Mon city in the state of Nagland in India, where three methods have compared, the first method have involved prediction using the (MLFF), while in the second one PCA has applied to the inputs of the MLFF network and prediction, and in the third one replacement the temperature series with the first PC have  proposed then (PCA) has applied to the inputs of the MLFF network and prediction. The results have showed the ability of the MLFF network to predict the weather parameters under study except the rainfall series as it is one of the series which doesn’t follow a specific rule, and the application of PCA approach to the inputs of the network has reduced the number of network parameters to be estimated, improved prediction results,  and the third proposed method has outperformed the Previous two methods for predicting the series under study.

KeyWords: Principal Components Analysis, Neural Networks, Prediction, Weather Parameters of Aleppo City, Weather Parameters of Mon City

Full paper in Arabic: PDF