ICACSIS 201-1zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Forecasting the Length of the Rainy Season Using Time Delay Neural Network Agus Buono, Muhammad Asyhar Agmalaro, and Amalia Fitranty AlmirazyxwvutsrqponmlkjihgfedcbaZYXW Department of Computer Science Faculty of Mathematics and Natural Sciences. Bogor Agricultural University Email: pudesha@gmail.com.agmalaro@gmail.com.lia.ilkom47@gmail.com Abstract-Indonesia has abundant natural resources in agriculture. Good agricultural resuIts can be obtained by deter mining a good growing season plan. One of important factors which determines the successful of crop is the length of the rainy season. The length of the rainy season is dynamic and difficuIt to be controlled. Therefore the planning of the growing season becomes inaccurate and cause crop failures. This research ai ms to develop a model to predict the length of the rainy season using time delay neural network (TDNN). Observattonal data used in this research is the length of rainy season from three weather and climate stations of the Pacitan region from 1982/1983 to 2011/2012. Predictor data used in this reserach is sea surface temperature (SST) derived from the region of Nino 1+2, Nino 3, Nino 4, and Nino 3.4 from 1982 to 2011. Model with the best accuracy was obtained by Pringkuku station with RMSE of 1.97 with pararneters of delay 10 1 2 3\, learning rate 0.1, 40 hidden neurons, and predictors of Nino 3 and R-squared of 0.82 with pararneters of delay 10 1\, learning rate 0.3, 5 hidden neurons, and predictors of Nino 3.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA I. I:-.rTRODUCTION lndonesia has abundant natural resources 111 agriculture. The quality of the agricultural products is influenced by a good planning of growing season. One important factor in a good planning of growing season is the lcngth of the rainy season which is diflicult to predict. Indonesia passed by west and east monsoon wind so lndonesia have two seasons, rainy and dry season Season in lndonesia is also influenced by global phcnomena as El Nino and La Nina II]. In tropical areas, El Nino and La , ina lead to a shift in rainfall pattems. change in the arnount of rainfall. and change the air tcmperature (2). This leads to the difficulty in predict the length of the rainy season 50 that the planning of growing season becomcs less precise and have an impact on crop failure. The lcngth of the rainy season is greatly affect rice production. especial\y in the second growing season. If the rainy season is short. then the chances of drought during the second growing increase and can cause the crop failure [3]. Information regarding the length of the rainy season is very useful for the parties involved in the good planning of growing sea son as much as possible in order to avoid crop failure and minimal losses. This study airns to solve the problem by building a good model in predicting the length of the rainy season. Prediction of the length of the rainy season in this study is using time delay neural network (TDNN) with SST as the predictor. TDNN is able to capture the diverse characteristics of the data [4) so it is suitable for the length of the rainy season data that diverse and uncertain, This method has extraction layer using a sliding window on the input layer 50 it is dynamic (4). Predictors used in this study are SST which is one of the global phenomenon that affects some variable rainfall and one of them is the length of the rainy season [5) and it was showed that there is strong correlation between SST with rainfall in lndonesia (6). The length of the rainy season data that used in this study is from Pacitan area. Based on data Pacitan from 1982 to 2009 indicated that about 90% of drought occured in the dry season (May. June. July. and August) and the rest occurred in the early rainy season (November and December), it is interesting because theoretically rain should be abundance in this period. th is indicates that there is a unique pattem in this data [7]. II. RESEARCH METHODS :I. Data Collection Data that used in the study were SST as a predictor and the length of the rainy season as an observation. SST data is obtaincd from the National Oceanic and Atmospheric Administration (:--iOAA).zyxwvutsrqponm liS Department of Agriculture from the region of Nino 1 ~ 2. :--J ino 3. , ino 3.4 and Nino 4 in each rnonth from 19R2 to 201 I. The length of the rainy season data on etimate and weathcr stations in each Pacitan region.