<![CDATA[Rainfall fluctuates with time and changes randomly, which unfavorable for most of the cropping, such as paddy. An early warning system is required to ensure a productive paddy cropping system. This paper describes the rainfall prediction modelling using a neural network analysis at paddy production centre area in the northern coast of Western Java and Banten. Rainfall data from Baros in the northern coast of Banten, Karawang, and Kasomalang Subang in the northern coast of West Java have been used for setting and validating the model. The model provides rainfall prediction for the next three months (Y=CHt+3), using the inputs data of the number of month (X1=t), the rainfall at the current month (X2=CHt), the rainfall atthe following month (X3=CHt+1), the rainfall at the following two months (X4=CHt+2), the southern ossilation index (SOI) at the current month (X5=SOIt) and the NINO-3,4 sea surface temperature anomaly at the current month (X6=AnoSSTt). Rainfall data recorded in the 1990-2002 period have been used for composing the model, and those in the 2003-2006 periods have been used for validating the model. The validated model has been used to predict rainfall in the 2007-2008. The best modelare those that using a combination of those six input variables. These models are able to explain 88-91% of the data variability with 4-8 mm month-1 of the maximum prediction error. At Baros Serang, the predicted rainfall in the 2007-2008 periods will be varied from Normal to Above Normal. At Karawang and Kasomalang Subang, predicted rainfall will be high at the end of 2007 until early 2008, and then will be low in the middle of 2008 and increases at the end of 2008.]]>
