Predicting Time Series Sales Transaction Value at E-Stall Crunchy Merchants Using Long-Short-Term Memory

The utilization of sales data as a foundation for strategic decision-making plays a crucial role in supporting the operational success of Merchant E-Stall Crunchy. However, the high variability and volume of daily transaction data create challenges in accurately identifying sales patterns. To address this issue, this research proposes the application of a deep learning approach—Long Short-Term Memory (LSTM)—to perform time series–based prediction of sales transaction values. The objective of this study is to develop a predictive model capable of uncovering historical patterns, seasonal trends, and demand fluctuations, thereby providing insights that support managerial decision-making. The Design Science Research methodology is employed, covering problem identification, model development, implementation, and evaluation. The dataset consists of daily sales transactions from E-Stall Crunchy, including transaction dates, product types, and sales values. The LSTM model is implemented to predict daily transaction values, and its performance is evaluated using MAE and RMSE metrics. The expected outcome includes an accurate prediction model and analytical recommendations that can support strategies for improving sales performance and operational efficiency.