<![CDATA[Wholesale apparel businesses need category-level forecasts to support stock planning, yet short and volatile sales series can make simple models unreliable. This study developed and evaluated a web-based sales forecasting system for Grosir 21.12 Store using ordinary least squares linear regression. A total of 4,063 transaction records from January 2024 to December 2025 were aggregated into 24 monthly observations for 15 clothing categories. Evaluation used a chronological split of 20 training months and four testing months, with a three-month simple moving average as the baseline. Across 14 categories with defined test MAPE values, linear regression produced a mean MAPE of 92.29%, compared with 62.98% for the moving-average baseline; regression was better in only two categories. The system nevertheless generated explicit category trends, automated forecasts, and role-based reports, while all 12 black-box scenarios produced the expected outputs. The findings show that the application is functionally feasible, but linear regression should be treated as a transparent trend baseline rather than an operationally accurate stock-forecasting model. Future evaluation should use longer series, seasonal and exogenous predictors, robust error measures, and prospective business outcomes.]]>
Copyrights © 2026
