RÉSUMÉ
A combined epidemic prediction method based on multi-source data fusion is presented to address the common problems of low accuracy,weak generalization,single structure,poor nonlinear processing ability,and long prediction time in traditional epidemic prediction models.The collected multi-source epidemic data are normalized and subjected to feature selection using principal component analysis.An ARIMA-GM-BPNN model for pandemic prediction is constructed by combining ARIMA model,grey GM model and BPNN.The fitting values of the first two prediction models are used as inputs to BPNN for model training.After sufficiently integrating the data and combining the advantages of different prediction models,the optimal combined model is obtained and used for forecasting the incidence and trend of epidemics.Experimental results show that the combined model exhibits excellent fitting performance,with predicted incidences and trends consistent with the real conditions.The proposed approach improves prediction accuracy and generalization capabilities,and it can provide reliable data support for epidemic prediction and control.