Online group streaming feature selection considering feature interaction

Abstract In real-world applications, features can be generated continuously one by one or by groups, such as image analysis and physical examination. Online streaming feature selection deals with streaming features on the fly. Existing streaming feature selection methods focus on removing irrelevant and redundant features and selecting the most relevant features, but they ignore the interaction between features. Interacting features appear to be irrelevant or weakly relevant to the class individually. However, if they are combined, they may highly correlate with the class. Features within the same group are more likely to interact with each other. Therefore, in this paper, we focus on feature interaction within and between the streaming groups and propose an Online Group Streaming Feature Selection method that can select Features to Interact with each other, named OGSFS-FI. OGSFS-FI consists of two stages: online intra-group selection and online inter-group selection. For intra-group selection, we design a new pair selection strategy that can select features interacting with each other. For inter-group selection, we use the regularization and variable selection method elastic net, which encourages a grouping effect. Extensive experiments conducted on synthetic and real-world datasets demonstrate our new method’s efficiency and effectiveness.