Figures & data
Figure 2. Transfer learned model from Inception-v3; the original Inception-v3 architecture graph (Main body) is adopted from Serengil (Citation2018).
![Figure 2. Transfer learned model from Inception-v3; the original Inception-v3 architecture graph (Main body) is adopted from Serengil (Citation2018).](/cms/asset/7824ce6e-4c75-4d50-b0d9-3fa397bf7a93/tjde_a_1633425_f0002_oc.jpg)
Table 1. Layer functionality summary.
Figure 5. Research area for two flooding cases with their geotagged tweets; (a) Continental U.S; (b) South Carolina flood in 2015 with 934,896 geotagged tweets from October 2 to October 9; (c) Houston flood in 2017 with 501,516 geotagged tweets from August 25 to September 1.
![Figure 5. Research area for two flooding cases with their geotagged tweets; (a) Continental U.S; (b) South Carolina flood in 2015 with 934,896 geotagged tweets from October 2 to October 9; (c) Houston flood in 2017 with 501,516 geotagged tweets from August 25 to September 1.](/cms/asset/63f4b0f9-f064-442c-b804-2d21c4460536/tjde_a_1633425_f0005_oc.jpg)
Table 2. Visual training set.
Table 3. Textual training set.
Figure 6. Model performance of the visual CNN: (a) the ROC curves for all 5 folds; (b) the training and validation accuracy curves for fold 3 (best fold) in two stages.
![Figure 6. Model performance of the visual CNN: (a) the ROC curves for all 5 folds; (b) the training and validation accuracy curves for fold 3 (best fold) in two stages.](/cms/asset/9525d361-8a95-43f2-9efd-19d3f387d573/tjde_a_1633425_f0006_oc.jpg)
Table 4. Visual CNN performance.
Table 5. Word2Vec training results (top 5 neighboring words with their cosine-similarity distances).
Figure 7. (a) Textual CNN ROC curve for all 5 folds; (b) Training accuracy curve for fold 1 during 200 epochs; (b) Training loss curve for fold 1 during 200 epochs.
![Figure 7. (a) Textual CNN ROC curve for all 5 folds; (b) Training accuracy curve for fold 1 during 200 epochs; (b) Training loss curve for fold 1 during 200 epochs.](/cms/asset/e8fc6266-06de-4fac-ad16-d67c24b8dc50/tjde_a_1633425_f0007_oc.jpg)
Table 6. Textual CNN performance.
Table 7. Visual-textual fused classification accuracy.
Table 8. Visual-textual fused classification compared with textual only.
Figure 8. ROC curves of the six algorithms using visual-textual fused vector () and using textual vector (
) alone; (a) LogR; (b) DT; (c) RF; (d) SVM (Linear); (e) SVM (RBF); (f) SVM (Sigmoid).
![Figure 8. ROC curves of the six algorithms using visual-textual fused vector (Zfused) and using textual vector (Ztextual) alone; (a) LogR; (b) DT; (c) RF; (d) SVM (Linear); (e) SVM (RBF); (f) SVM (Sigmoid).](/cms/asset/feae94bc-9a32-48f4-b99b-2b074332efa4/tjde_a_1633425_f0008_oc.jpg)
Figure 9. Eight Examples of classification results. denotes the probability of a post being flood relevant. Blurring was applied to faces appeared in the photo.
![Figure 9. Eight Examples of classification results. Pflood denotes the probability of a post being flood relevant. Blurring was applied to faces appeared in the photo.](/cms/asset/882d270b-0943-44fe-a931-df461de9010e/tjde_a_1633425_f0009_oc.jpg)