<![CDATA[Daily order fluctuations make it difficult for motorcycle ride-hailing drivers to plan online working time. This study develops an interpretable multiple linear regression model to predict the number of daily GoRide orders from online hours, day type, and weather. The dataset contains 100 driver-day observations collected from eight GoRide driver-partners operating in central Tulungagung. To preserve observation order, the first 70 records were used for training and the last 30 for testing. Day type and weather were expressed as binary indicators, yielding the operational equation ŷ = −1.9887 + 1.0036X₁ + 1.4187Dweekend + 2.3968Dclear. The model obtained R² = 0.8440 and MAPE = 15.62% on training data. On the held-out test set, it achieved R² = 0.8341, MAPE = 13.48%, MAE = 0.989 orders, and RMSE = 1.322 orders. Twenty of 30 predictions (66.7%) were within one order of the actual value, and 26 (86.7%) were within two orders. However, predictions were positively biased by 0.784 orders on average, with 23 of 30 observations overpredicted and a maximum absolute error of 4.458 orders. The model was integrated into a Flask–MySQL web application for data management and interactive prediction. The results show that a compact linear model can provide useful and explainable driver-level estimates, but its reliability remains conditional on the small, locally collected sample. Rolling validation, driver-aware evaluation, richer temporal variables, and count-regression benchmarks are required before operational deployment at scale.]]>
Copyrights © 2026
