Analisa Metode Backpropagation Pada Prediksi Rata-rata Harga Beras Bulanan Di Tingkat Penggilingan Menurut Kualitas

Abstract

Rice is a staple food in Indonesia and plays a crucial role in the food structure as a source of nutrition. The diverse population of Indonesia, spread across various islands, makes rice availability highly important. The government continues to strive for food security, particularly by increasing domestic production. These considerations become even more significant for Indonesia due to its growing population and extensive geographical distribution. To meet the population's food needs, Indonesia needs sufficient food supply and distribution to fulfill consumption and maintain adequate reserves for extensive logistical operations. Rice shortage can be seen as a threat to economic and political stability. The significance of rice as a food commodity means that it is constantly in demand by people from all walks of life. Price fluctuations over time due to imbalances between supply and demand have a significant impact on the middle class and working class. The instability of rice prices greatly affects both the general public and farmers. Generally, prices are determined by the interaction of supply and demand. If supply is high and demand is low, prices will decrease. Conversely, if supply is low and demand is high, prices will increase. Prediction is an important tool to anticipate future events by recognizing patterns from the past. Backpropagation can be used as a method to predict rice prices. The data used in this study are the average monthly rice prices at the milling level according to the quality of large-scale traders from January 2023 to December 2023, in Indonesian Rupiah per kilogram. This research utilizes data obtained from the website of the Indonesian Central Bureau of Statistics (BPS) from 2013 to 2022. The study employed 5 different architectures for data testing, namely the 15-15-1 architecture with a testing mean square error (MSE) of 0.00644604, the 15-19-1 architecture with a testing MSE of 0.01005532, the 15-30-1 architecture with a testing MSE of 0.02119922, the 15-31-1 architecture with a testing MSE of 0.00009938. The best architecture in this study was the 15-17-1 model with 5206 iterations and a runtime of 18 seconds, achieving the smallest testing MSE of 0.00000105 and the highest accuracy of 100%. From the obtained architectures, it is evident that backpropagation can perform with a high level of precision. This research can serve as a guideline for the government to determine rice availability and establish average rice prices based on quality, thus preventing future rice shortages.