<![CDATA[Manual sales processes and stock management in micro, small, and medium enterprise (MSME) settings often lead to limited market reach and inefficient inventory management. The purpose of this research is to solve these operational problems by designing and developing a web-based e-commerce system equipped with an integrated monthly stock demand forecasting module to enhance the competitiveness of MSMEs in the roster industry. The system was developed following the waterfall methodology and it has a decoupled architecture to separate the artificial intelligence computational workloads from the core application. Two time series forecasting models Long Short-Term Memory (LSTM) and Facebook Prophet were applied and compared for forecasting stock requirements from intermittent, zero-inflated demand patterns of historical sales data. System functionality was validated using User Acceptance Testing and the forecasting accuracy was measured using Root Mean Square Error (RMSE) and Weighted Mean Absolute Percentage Error (WMAPE). The performance evaluation showed that the unscaled LSTM model outperformed the linear additive regression method of Facebook Prophet in terms of lower physical volume deviation and consistent operational error in the course of the evaluation period. The developed platform provides a reliable data-driven decision support for inventory management. The incorporation of forecasting using neural networks in the e-commerce system has reduced the risk of stockouts, expanded the market, and proved the increase in the business competitiveness of artisan MSMEs.]]>
