Prediction of rainfall by using nearest neighborhood approach / Muhammad Shofwatullah Mohamed Radzwan, Mohamad Nazran Ab Ghani and Mohamad Amirul Aiman Che Abdullah

Prediction of rainfall has become a normal phenomenon in the world, usually people will think that predict rainfall is normal and just like predict it without any calculation. However, for some knowledgeable people and researchers, they know that the prediction of rainfall can be used by usmg the mathematical calculation. Therefore, in our research, we study about one of them, which is the nearest neighborhood approach. Therefore, the aim of this study is to find the suitable model of nearest neighborhood approaches which are zero-order approximation method (ZOAM) and improved ZOAM. The improved ZOAM are k-nearest neighborhood approximation method (KNNAM), and weighted distance approximation method (WDAM). Moreover, to compare the performance of accuracy between the nearest neighborhood approaches by finding the relationship between the real and predicted data by using correlation coefficient, cc and use the average absolute error, e to decide whether the model is better or not. Lastly, to show whether short or long term which one is better as a length of prediction range. Based on this study, the result shows that the most suitable model to predict the rainfall data is WDAM by using 5 distances, and the best duration of prediction is predicting in the short-term by using WDAM with 8 distances. Overall, WDAM is the best model because add-on the element of distance to improve the prediction tool and better than ZOAM and KNNAM.