SISTEM PENDUKUNG KEPUTUSAN BERBASIS JARINGAN SYARAF TIRUAN UNTUK PERAMALAN HARGA KOMODITAS TANAMAN PANGAN

• Main Authors: , Ferlando Jubelito Simanungkalit, , Dr. Ir. Lilik Sutiarso, M.Eng
• Format: Theses and Dissertations NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2012
• Subjects: ETD
• Online Access: https://repository.ugm.ac.id/99071/; http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=54850
• Institution: Universitas Gadjah Mada

Decision Support System Based on Artificial Neural Networks For Food Crop ABSTRACT Commodities Price Forecasting was designed to provide a stimulus for decision makers concerning food price stabilization, future price trend and available planting schedule policies which enable to maximize the profit. The main purpose of this study was making the design of Decision Support System (DSS) by firstly analyzing the architecture of Artificial Neural Networks (ANN) that appropriate to be used as forecasting method/model base of the DSS. The study was done by using the monthly prices of the food crop commodities in Sleman Regency, D.I. Yogyakarta province, from January 2000 to July 2011. The best architecture was selected based on the lowest value of Mean Square Error (MSE) and Mean Absolute Percentage Error (MAPE) from training, testing and validation result. Then, the best architecture was designed to be the model base of the DSS as well as the database, user interface and elements of knowledge by using the decision support system developing phases and programmed with the programming language. From the 324 trials unit of the ANN architecture analysis for each commodity, it has been obtained that there was a best ANN architecture for each commodity and valid to be used as the forecasting method with 15% tolerance of MAPE. From 6 varieties of food crop as the object of study, the very best ANN architecture derived from rice IR64 with the architecture [12 � 32 � 1], learning rate 1,75 and the transformation range of the data [0 and 1], with consecutive value of MSE and MAPE in training, testing and validation process was [0,00125 and 2,807%], [0,0219 and 3,289%], [0,0244 and 3,575%]. Based on the validation result, the limit of the forecasting period that still valid to be done by the system was in the next 12 months. The result of the study was the ANN architecture used by the system met to the preformance degradation in terms of the price pattern which fluctuating sharply, it was because the ANN architecture used by the system was not considered some factors that could make the fluctuation of price, therefore the development of the ANN architecture was needed as the model base of the DSS in order to improve the ability of the system to provide the better decision support.