Penerapan Model Long Short-Term Memory (LSTM) dalam Prediksi Pendapatan dan Stok Berbasis Deret Waktu pada UMKM

Abstract

Revenue fluctuations in egg wholesale businesses are generally influenced by changes in market prices, consumer demand levels, and seasonal factors, which often make it difficult for business owners to conduct accurate financial planning. Using time series data, this study aims to predict the revenue of Rizki Egg Wholesale Store by applying the Long Short-Term Memory (LSTM) method. LSTM is selected due to its capability to capture long-term patterns in nonlinear data and to produce more reliable predictions compared to conventional methods such as linear regression. The research data include gross revenue and operational costs, which are used to calculate net profit as the basis for analysis. The research process involves data collection, preprocessing using MinMaxScaler, development and training of the LSTM model, and evaluation using the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) metrics. The findings indicate that the LSTM model produces prediction errors with MAPE values of 11.82% and 5.00%. A MAPE value of 5.00% is classified as very good (<10%), while a MAPE value of 11.82% falls into the good category (10–20%). Therefore, the LSTM model is considered sufficiently effective in predicting revenue and has the potential to support more efficient business decision-making.