The Internet of Things (IoT) is expanding rapidly due to the fast development in information and communication technology. It refers to the collective networks of connected devices that simplify the communication between devices and clouds, as well as between the devices themselves. Due to the dynamic environment of IoT devices, static training models to make predictions are inefficient. A model with incremental training is sufficient to handle the dynamic nature of IoT predictions.
Incremental learning is a suitable learning model for IoT prediction, which can learn new data sequentially without forgetting the existing knowledge. Changes in the environment of IoT predictions differ for every class, and it is better to use incremental learning with respect to class for IoT predictions. Class incremental learning focuses on the learning classification model with the number of classes increasing stage by stage. Class incremental learning models produce significantly accurate predictions for large-scale IoT applications.