<![CDATA[Apotek XYZ faces significant challenges in drug stock management due to unpredictable seasonal demand fluctuations, particularly during flu season, which risks stock shortages or excess inventory. This study implements a Long Short-Term Memory (LSTM) method for time-series-based drug sales forecasting and develops the "Riycast" web dashboard as an interactive stock management solution. Historical daily sales data (January 2021–December 2024) for 10 key drugs (e.g., multivitamins, flu medications) were processed through CRISP-DM stages including data cleansing, normalization, seasonal decomposition, and hyperparameter tuning via grid search. The LSTM model captured seasonal patterns and trends with variable accuracy (RMSE 0.11279–0.31552), peaking for Ultraflu and Vitalong Z Sinc. The Riycast dashboard built with Flask(backend), React.js (frontend), and MySQL features real-time sales data input, interactive prediction visualizations, historical trend analysis, and automatic surge alerts (>100 units). Implementation boosted stock management efficiency by 30% in trials, reduced stockout risk by 25%, and enabled data-driven decisionmaking at Apotek XYZ.]]>
Copyrights © 2026
