Three-order improvement early-warning model for structural safety of reconstructed houses
In order to study the structural safety problem of reconstructed houses and improve the early-warning accuracy of the model,three methods of"text-image fusion","information regeneration"and"intelligent optimization model parameters"were used to establish a three-order improvement early-warning model.Firstly,the basic early-warning model was constructed,then four image recognition models such as VGG16 and ResNet50 were selected for transfer learning,and the model with the best performance was used as the basic early-warning model.After that,the first early-warning accuracy improvement was car-ried out.The text information corresponding to the test set was collected,pre-processed by one-hot coding and"fused"with image information,and five machine learning algorithms such as random forest were selected to improve the early-warning ac-curacy.Then,the second accuracy improvement was carried out.The"oversampling-deep convolutional generative adversarial network"strategy was used to improve the"capture"ability of the model to the reconstructed houses with potential safety hazard.Finally,the golden jackal optimization algorithm was used for the third improvement.The results show that the DenseNet121 model can better capture the image features of the reconstructed houses with potential safety hazard.The optimal early-warning model for the structural safety of reconstructed houses is SVM,with the accuracy of 82.5%.After using the"oversampling-deep convolutional generative adversarial network"strategy,the SVM and XGBoost with the best performance increase the recall rate of the reconstructed houses with potential safety hazard by 10 and 5 percentage points,respectively.The overall accuracy,recall rate,precision rate and F1 value of"SMOTE-DCGAN-SVM"under the golden jackal optimi-zation algorithm increase by 7.0,7.5,10.5 and 9.1 percentage points.The research results can provide a reference for the relevant departments to infer the reconstructed houses that may have structural safety hazards in advance when conducting the census of dangerous houses.
NETL
NSTL
万方数据
