Aiming at the problems of insufficient classification accuracy of production material inventory and difficulty in demand fluctuation prediction of medical device manufacturing enterprises under the background of medical demand growth, this paper takes the medical device inventory in the smart medical supply chain as the research object, and constructs an intelligent inventory identification and prediction model combining ABC-CVA classification and CNN-LSTM deep learning method. The model classifies and identifies inventory materials from two dimensions of capital occupation and production criticality, and then combines the local feature extraction ability of CNN and the time series dependence modeling ability of LSTM to predict the dynamic inventory level of different types of materials. The results show that the overall prediction performance of the constructed CNN-LSTM model is better than that of the baseline model, and it can meet the prediction management requirements of class A high-value materials and CVA critical materials. Under the monthly prediction task, the mean absolute error, root mean square error and symmetric mean absolute percentage error of class B materials are reduced by at least 2.12%, 2.75% and 1.88%, respectively, and the mean absolute error, root mean square error and symmetric mean absolute percentage error of class C materials are reduced by at least 4.15%, 7.38% and 3.65%, respectively. The model can provide data support for medical device manufacturing enterprises to carry out inventory forecasting, dynamic early warning, resource allocation and replenishment decision-making.
