Supply chain inventory management faces the problems of inaccurate demand prediction and large inventory fluctuation, and the accurate prediction and optimization method based on spatio-temporal data mining can effectively improve the operational efficiency and decision-making quality. This study constructs a supply chain inventory demand forecasting model and optimizes inventory management through spatio-temporal data mining methods. The study adopts ARIMA model for inventory demand forecasting and combines the system dynamics method to establish the supply chain inventory optimization model. Based on the historical inventory data of a Guangzhou food company (Company A) from January 2012 to December 2023, the data from January 2012 to June 2019 are used as the training set, and the data from July to December 2023 are used as the test set for empirical analysis. The optimal forecasting model is identified as ARIMA(0,1,1) through the series smoothing test, white noise test and model order fixing. The results show that the ARIMA(0,1,1) model performs better in forecasting the first quarter of 2026 with a MAD value of 167 and a MAPE value of 5% compared to the Winters multiplicative model, which has a MAD value of 461 and a MAPE value of 9%. Based on the demand forecasting results, a twolevel supply chain (supplier and retailer) system dynamics model was constructed, containing 10 constant parameters and 27 dynamic variables. The simulation analysis was carried out by VENSIM software for a 50-day cycle, and the optimized model showed that the fluctuation of the inventory curve was reduced, the order quantity decision was more accurate, and the value of the unsatisfied demand was greatly reduced and smoother. The conclusion of the study shows that the demand forecasting and inventory optimization method based on spatiotemporal data mining can effectively reduce the inventory risk and improve the efficiency of supply chain operation, and it is suggested that enterprises should strengthen the construction of the information sharing mechanism, and shorten the supply chain lead time through the optimization of the business process and the delaying strategy.
