A survey and analysis of feature selection techniques in machine learning for IoT device classification within smart buildings

The Internet of Things (IoT) has revolutionized modern living and infrastructure by driving the development of sustainable smart buildings and accelerating the digital transformation of buildings. In smart buildings, efficient machine learning (ML) based classification of IoT devices is critical for improving cyber defense, optimizing resource management, and maintaining occupant comfort. Feature selection techniques are vital for boosting the effectiveness of machine learning models when classifying and categorizing Internet of Things (IoT) devices for various reasons. Hence, this study initially provides an in-depth understanding of integrating IoT and ML in smart buildings. We provide the reasons and importance of device classification in smart buildings, which may range from monitoring security, power consumption, resource allocation, maintenance, and rehabilitation scenarios. This study emphasizes the importance of feature selection (FS) models in enhancing the accuracy of classification and interpretability for diagnosing and managing smart building systems effectively. This study thoroughly provides the state of the art for feature selection techniques in detail, with their purpose. It evaluates the principles and the types of feature selection methods, including their applications. It also highlights the key issues and challenges faced in applying these techniques in smart building infrastructures. This study discusses the process of optimization of feature selection methods, which helps to improve the model’s effectiveness and speed up machine learning accuracy for secure smart building resilient structures, for its various benefits. Lastly, we provide a detailed discussion and suggestions along with future perspectives of FS in ML for IoT device classification within smart buildings.