Prediksi Produksi Tanaman Padi di Sumatera Dengan Menggunakan Algoritma Neural Network

Abstract

Rice production is a crucial factor for food security in Indonesia. This study aims to predict rice production in Sumatra using the Neural Network algorithm implemented through RapidMiner. The data used includes variables such as year, production, harvested area, rainfall, humidity, and average temperature, collected from eight provinces in Sumatra for the period 1993-2020. The research process involved dividing the dataset into training data (90%) and testing data (10%), normalizing data using the proportion transformation method, and training the model with parameters such as 650 training cycles, 0.9 learning rate, and 0.3 momentum. Model evaluation utilized root_mean_squared_error (RMSE), yielding an RMSE value of 0.006, indicating high prediction accuracy. The results demonstrate that the model is effective in predicting rice production, particularly for data with normal distribution. Recommendations for further development include creating a web-based visual interface to support end-users and applying hyperparameter optimization to enhance model performance. This model is expected to serve as a decision-support tool for strategic planning in the agricultural sector.