MIS-IOT: Modular intelligent server based internet of things framework

In this thesis, the aim is to develop modular and open source internet of things framework. The framework is designed with three separate modules as front end module, clustered server module and learning module. Since the framework is designed with modular approach, adding new features and maintaining the each module is much easier. Each module can be deployed to different host separately and communicate with using REST architecture. The front end module is responsible for offering all framework features with user interface, thus providing better user experience. The server module is responsible for managing all framework features. The learning module handles anomaly analysis of time series data with using LSTM algorithm. The user can upload new data as CSV(Comma Separated Values). In addition to bulk loading data, the user can also create socket or REST connection in desired port for listening data from various sources. Data flow on specific port can also be tracked with user interface. After uploading or listening data from various data sources, learning tasks can be created. The framework supports three different learning task types as task with framework data, task without framework data and LSTM task. Through using task with framework data option, user can create learning task with using existing data with specific criteria. Data is converted to JSON and passed as an argument to the desired Python script. Through using task without framework data option, learning task is created without using any framework data and specified script is executed directly. In LSTM option, the learning task can be created with specifying LSTM parameters and data. In this option, anomaly analysis is made on specified dataset. Result of the anomaly analysis can be seen on the graph using the front end module.