Research on Intelligent Scheduling Application for Raw Material Warehousing Based on Transformer Neural Network

With the rapid development of China's tobacco industry, raw material warehousing scheduling faces challenges such as rough inventory management, delayed response, and insufficient informatization. To improve resource utilization and reduce costs, this paper proposes an intelligent scheduling model based on a Transformer neural network. The model is compared with a traditional Logistic regression model. Data preprocessing ensures data completeness and quality through missing value imputation and normalization. Logistic regression uses a Sigmoid function to map a linear combination to event occurrence probabilities. Although its structure is simple and easy to interpret, it has limitations in handling complex data. To overcome these limitations, the Transformer model utilizes self-attention, multi-head attention, and positional encoding. It captures global features and potential associations. It achieves precise prediction of warehousing scheduling demands. Experimental results show that the Transformer significantly outperforms the traditional model in terms of accuracy, recall, F1-score, and AUC. It also performs better on error metrics such as MAE and RMSE. These findings demonstrate its advantages in mining nonlinear relationships and modeling long time-series data. This study not only provides an innovative data-driven solution for tobacco raw material warehousing scheduling, but also offers theoretical support and experimental evidence for the application of deep learning technology in industrial scheduling.