Tree Preservation Order (TPO) is used to protect specific trees from damage and destruction, which is determined in high subjectivity. This research collected and analyzed TPO data, aerial images, geographic data, and socio-economic data in the Greater London area and developed a multi-input deep learning (DL) framework to classify TPO-protected and non-TPO-protected trees. The synergy use of aerial images and GIS data with the fusion model of ResNet50 and multilayer perceptron network produced the best classification accuracy of 87.32%. The result indicated the robustness of the multi-input DL model to identify the social attributes of trees compared with the single-input DL model