Inference and Prediction
Inference and Prediction functional component is responsible for providing high-level knowledge from raw IoT data using different pattern recognition techniques. In this context, we have explored several supervised machine learning techniques including different variants of Support Vector Machines and K- Nearest Neighbor, in addition to statistical techniques such as Hidden Markov Model (HMM). In short, it provides the following two main functionalities:
1) If labelled historical data is available (raw data with labelled high-level knowledge), it provides the functionality to train the model and provides capability to deploy the model in order to predict the output for real-time data;
2) If the labelled historical data is not available (incomplete data), it exploits the temporal patterns of the data and learns using statistical properties of the data to train the model which can be used to predict the output for real-time data.
More information can be found in Deliverable D6.1.2