Address unit analysis method and device based on multi-model weighted fusion

The invention provides an address unit analysis method and device based on multi-model weighted fusion, and belongs to the field of address unit analysis and the like, and the method comprises the following steps: S1, training a baseline model marked by an address unit; s2, training a non-baseline model used for address unit labeling; s3, constructing a fusion model; and S4, labeling the address units by adopting the fusion model. And calculating a weight according to the difference between the F1 values of the to-be-fused model and the baseline model, wherein the weight is used for fusing the predicted label scores of different models. When the fusion weight of the model is calculated, the F1 value is adopted to measure the quality of the model, which is more scientific and comprehensive and has higher reliability than the accuracy to measure the quality of the model; the weight during model fusion is calculated according to the difference between the F1 values of the to-be-fused model and the baseline model