A GRU-attention-based method for filling missing values in agricultural drying process monitoring time series data

In the process of tobacco drying, the integrity of time series data is crucial for ensuring product quality and production efficiency. However, due to sensor failures, environmental interference, or data transmission issues, missing values in the data are inevitable, severely affecting the accuracy and reliability of data analysis. Therefore, to improve the quality of monitoring data, this paper proposes a GRU-Attention-based method for filling missing values in time series data during the tobacco drying process. This model incorporates an attention mechanism and skip connections to optimize the imputation process. Additionally, it introduces a forward and backward imputation fusion method combining confidence and time decay, and enhances the accuracy of imputed values through iterative smooth moving imputation. Applying this method to fill missing values in tobacco drying monitoring time series data, experimental results demonstrate that it effectively imputes missing values, thereby improving the data quality of tobacco drying monitoring. Compared to other imputation methods, this approach offers higher imputation accuracy.