A Rainfall Forecasting Method Using Machine Learning Models and Its Application to Fukuoka City Case

In the present article, an attempt has been made to derive optimal data-driven machine learning methods for forecasting average daily and monthly rainfall of Fukuoka city in Japan. This comparative study has been conducted from three aspects: modelling inputs, modelling methods and pre-processing techniques. A comparison between linear correlation analysis and average mutual information is done to find the optimal input technique. For modelling of the rainfall, a novel hybrid multi-model method is proposed and compared with its constituent models. The models are, 1) artificial neural network, 2) multivariate adaptive regression spline, 3) k-nearest neighbour, and 4) radial basis support vector regression. Each of the above methods are applied to model the daily and monthly rainfall, coupled with a pre-processing technique including moving average and principal component analysis. In the first stage of the hybrid method, sub-models from each of the above methods are constructed with different parameter settings. In the second stage, the sub-models are ranked by a variable selection technique and the higher ranked models are selected based on the leave-one-out cross-validation error. The forecasting of hybrid model is done by the weighted combination of the finally selected models.