The Internet of Things (IoT) activates massive data flow in the real world. Each computer can presently be linked to the internet and supply useful decision-making information. Virtually sensors are implemented in every aspect of life. From different sources of sensors can produce raw data. Due to the various data sources, the method of extracting information from the flow of data is mostly complicated, networks inadequate and criteria for real-time processing. In addition, an issue of context-aware data processing and architecture also present, despite the fact that they are essential criteria for stronger IoT structure. In order to meet this issue, we recommend a Context-aware Internet of Things Middleware (CAIM) architecture. This enables the incorporation of highly diverse IoT application context information by using light weigh protocol MQTT (Message Queue Telemetry Transport) for transmitting basic data streams from sensors to middleware and applications. In this paper, we propose a contextualization which means that obtain data from sensors of different sources. First have to create a context profile with the help of context type like user, activity, physical, and environment context. Then also is create a profile by using attributes. Finally, raw data can be change into contextualized data through CAPS (context-aware Publish-Subscribe) hybrid approach. This paper discusses the current context analysis strategies that use either rational models or probabilistic methods exclusively. The evaluation of identifying contextualization methods shows the shortcomings of IoT sensor data processing as well as offers alternative ways of identifying the context