Incremental one-class classifier based on convex-concave hull

One subject that has been considered less is a binary classification on data streams with concept drifting in which only information of one class (target class) is available for learning. Well-known methods such as SVDD and convex hull have tried to find the enclosed boundary around target class, but their high complexity makes them unsuitable for large data sets and also online tasks. This paper presents a novel online one-class classifier adapted to the streaming data. Considering time complexity, an incremental convex-concave hull classification method, called ICCHC, is proposed which can significantly reduce the computational time and expand the target class boundary. Also, it can be adapted to the gradual concept drift. Evaluations have been conducted on seventeen real-world data sets by hold-out validation. Also, noise analysis has been carried out. The results of the experiments have been compared with the state-of-the-art methods, which show the superiority of ICCHC regarding the accuracy, precision, and recall metrics.