Artificial Neural Network for Weather Forecasting: A Review

Forecasting is considered one of the most imperative and perplex proceedings in the present-day. Weather forecasting is a natural process which entails the prediction of the change in the atmospheric conditions with the passage of time. It has become a core area of studies and analysis for the scientist owing to the brusque forecasts and approaches of weather. Weather forecasting has always been a challenging task. In several cases, scientists and researchers had attempted in establishing a linear association between the input and targeted data of the weather. Since, weather is non-linear and dynamic in nature, the target has shifted to prediction of non-linear weather data. Though weather prediction is relatively a statistical measure and is automated, with the traditional tools, its result is rather uncertain and not always accurate. Due to its non-linearity and complex process, the best approach for resolving such problems is with the use of Artificial Neural Network (ANN). ANN simplify the weather predictions with its better efficiency, reliability and accuracy. The features of ANN are not just to analyse the past data but also to acquire future predictions rendering to be much ideal for weather forecasting. NN is rather a complicated network which are pliable and flexible in nature. It is autodidactic with its existing training data consequently forging a new smart pattern useful for predicting the weather. Survey of various techniques of NN for weather prediction is provided and it is observed that by simply increasing the number of hidden layers the trained NN can predict and classify the weather variables with minimal error. In this paper, predictive analysis algorithm is incorporated with back propagation network(BPN) to predict future weather by training the network. The technical milestone, where various researchers have acquired on this discipline has been reviewed and presented in these surveys.