An Embedded Co-AdaBoost based construction of software document relation coupled resource spaces for cyber-physical society

Software is a very important means of achieving the vision of the cyber-physical society. Software document relation coupled Resource Spaces prompts the cyber-physical society by facilitating the reuse of software design knowledge. The establishment of software document relation coupled Resource Spaces faces the scarcity of labeled data that helps discovering software document relations between resources dwelling in different Resource Spaces. This paper proposes the Embedded Co-AdaBoost algorithm to overcome this challenge by making the best use of easily available unlabeled data, integrating multi-view learning into the AdaBoost and leveraging the advantages of Co-training for performance enhancement. Compared with conventional AdaBoost, the experiment illustrates the effectiveness of the Embedded Co-AdaBoost in the convergence rate, the accuracy and the steady performance. The empirical experience demonstrates the ability of the Embedded Co-AdaBoost in prompting the development of software document relation coupled Resource Spaces.