<![CDATA[In the automotive industry, sales prediction is an important element in supporting strategic decision making, including production planning and supply chain management. This research focuses on applying the Gated Recurrent Unit (GRU) method to predict Toyota car sales in Indonesia based on monthly data from January 2011 to May 2023. GRU, as part of an artificial neural network, offers advantages in handling complex non-linear patterns in time series data. However, GRU also has challenges such as the risk of overfitting and the need for complex parameter tuning. Therefore, in this study, various regularization techniques such as dropout, batch normalization, L1 and L2 penalties, and early stopping are applied to improve model generalization. The evaluation results showed that the application of regularization techniques significantly improved the model performance, with a decrease in MAE by 20.9%, MSE by 23.1%, RMSE by 12.3%, and MAPE by 18.5% compared to the initial model. These findings suggest that the combination of GRU with regularization techniques can be an effective approach for sales prediction in the context of time series data.]]>
Copyrights © 2025
