Electronic Theses and Dissertation

Harga gula pasir di kota banda aceh, sebagai komoditas pangan strategis, mengalami fluktuasi yang berdampak pada stabilitas ekonomi dan ketahanan pangan. penelitian ini bertujuan meramalkan harga gula pasir menggunakan metode long short-term memory (lstm) untuk mendukung perencanaan logistik dan pengambilan kebijakan berbasis data. data harga harian dari pusat informasi harga pangan strategis nasional (pihps) periode 3 januari 2022 hingga 31 maret 2025, sebanyak 1.250 observasi, diolah melalui prapemrosesan, normalisasi dengan min-max scaling, dan pembagian data pelatihan-pengujian (80:20). model lstm dirancang dengan tiga lapisan (100-100-50 unit), dioptimalkan menggunakan algoritma adam, dan dievaluasi dengan metrik root mean squared error (rmse), mean absolute error (mae), dan mean squared error (mse). hasil evaluasi menunjukkan performa unggul dengan rmse=0.13, mae=0.10, dan mse=0.02. peramalan jangka pendek (18–30 april 2025) menghasilkan harga rp16.300–rp16.800, sedangkan proyeksi jangka panjang (april 2025–april 2026) memprediksi kenaikan hingga rp19.366, mencerminkan pola musiman. model diintegrasikan ke dalam dashboard interaktif berbasis streamlit untuk visualisasi real-time, memudahkan analisis oleh pemangku kepentingan. penelitian ini merekomendasikan pengembangan model dengan memasukkan variabel seperti inflasi dan kebijakan impor serta penerapan pada komoditas pangan lain untuk memperkuat sistem prediksi harga. kata kunci: peramalan harga, long short-term memory, gula pasir, deret waktu, dashboard, ketahanan pangan.

Price volatility of refined sugar in Banda Aceh, a strategic food commodity, impacts economic stability and food security. This study aims to forecast sugar prices using the Long Short-Term Memory (LSTM) method to support data-driven logistics planning and policy-making. Daily price data from the National Strategic Food Price Information Center (PIHPS), spanning January 3, 2022, to April 17, 2025 (1,250 observations), were processed through preprocessing, Min-Max Scaling normalization, and an 80:20 training-testing data split. The LSTM model, designed with three layers (100-100-50 units) and optimized using the Adam algorithm, was evaluated with Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Squared Error (MSE). The model achieved robust performance with RMSE=0.13, MAE=0.10, and MSE=0.02. Short-term forecasts (April 18–30, 2025) predict prices ranging from IDR 16,300 to IDR 16,800, while long-term projections (April 2025–April 2026) indicate a rise to IDR 19,366, reflecting seasonal patterns. The model was integrated into a Streamlit-based interactive dashboard for real-time visualization, facilitating stakeholder analysis. The study recommends enhancing the model by incorporating external variables such as inflation and import policies and extending its application to other food commodities to strengthen price forecasting systems. Keyword: Price forecasting, Long Short-Term Memory, refined sugar, time series, dashboard, food security.