Selecting Embedded Feature Modeling with Software Product Line for Smart Home Applications

Internet of Things (IoT) for environmental applications such as smart homes, campuses, cities, and health care systems are used to connect various wireless devices. Despite the use of the same devices, the reusability of these applications enables developers and software engineers to make enhancements to the development process according to the requirements of end users. However, the large number of variations in IoT technologies makes it hard for the developers or software engineers to select the default features in each application because of compatibility issues among devices. In this paper, Software Product Line (SPL) approach is proposed to design and manage the variabilities and commonalities among smart home applications. An embedded feature selection mechanism is proposed to classify the services by selecting the optimal subset feature. The selected subset features are those that are highly correlated and consistent with other features. The dataset is generated using open smart home simulator (OpenSHS) for daily activities within the virtual environment. The subset of features is evaluated by the level of consistency in each service. Weka data mining tool is used to find the classifier such as Bayes network and decision table which achieve the best accuracies on the dataset obtained in order to support services for residents in smart home applications. The approach adopted is based on a set of predefined performance metrics in order to compare the results for each classifier in terms of its ability to classify the services. It has been found that the Bayes network achieves an accuracy over 84.0% while the decision table achieved an accuracy over 83.0%.