Implementation And Analysis Of Arima Model And Kalman Filter For Weather Forcasting in Spark Computing Environment

In this paper, Role of SPARK in weather Forecasting is analysed by doing a comparative analysis for weather forecasting with and without the use of Spark cluster for ARIMA model and Kalman Filter. The values predicted by the above models is given to XGBoost Classifier to classify the weather condition. Last 20 years of data has been chosen from kaggle. Initially pre-processing is done and three attributes are narrowed down, namely humidity, temperature and dew point along with its timestamp. Weather Forecasting is done using the above models followed by XGBoost based on these three attributes. The predicted values of weather forecasting have been compared with actual values in the dataset for determining the quality of the models. To ensure the quality of prediction by the model in real-time, second order differencing method is introduced in Kalman filter for forecasting. The graphical analysis has been performed to analyse the computational time and quality of the models.