Machine learning methods have achieved human-level accuracies in many computer vision and natural language processing tasks. These techniques have led to advances in not only medical imaging, gaming and robotics but also in urban analytics. Previous research has begun to apply these learning methods to estimate socio-economic indicators using urban imagery. However, limited research studied how different urban form data can be combined to improve its performance. The aims of this research is to test and explore the efficacy on combining three sources of urban data to make inferences on socio-economic, transport and environmental indicators for the case study of Greater London, UK.