The Internet of Things is currently one of the most representative sources of Big Data. It can acquire real-time data from multiple spatially distributed points, allowing for the extraction of valuable insights. However, an appropriate integration, processing, and analysis of these data depends on several factors starting from the correct definition of the information systems. This paper introduces STS4IoT, a UML profile and automatic code-generation tool for model-driven IoT, to address this issue. STS4IoT allows designing and implementing an IoT application from the required data only, bridging the gaps between the IoT and database design worlds. The IoT data design includes both different in-network transformations and the join of streams from multiple sources. Besides, it follows the Model-Driven Architecture (MDA) guidelines to provide abstraction levels oriented to the different roles participating in the application design. The STS4IoT validation shows it has an excellent structure and is highly understandable. Its instance models are well-formed, highly abstract and readable. And the automatic implementation tool can generate complete code for complex real-world applications involving multiple IoT devices. Then, STS4IoT simplifies the definition and development of IoT applications and their integration into other information systems, such as stream data warehouses.