A Novel Imperialist Competitive Algorithm for Scheme Configuration Rules Mining of Product Service System

When designing a product service system scheme, the most critical issue is how to quickly and accurately translate customer value requirements into scheme technical characteristics. It is difficult to solve this problem by only the experience and knowledge of engineers. Effective configuration rule is necessary to assist engineers in scheme design. Therefore, a method of scheme configuration rules mining of product service system based on objective data rather than subjective data is proposed, and a new imperialist competition algorithm is introduced as an alternative method to solve multi-objective optimization problems. On the basis of the traditional imperialist competition algorithm, the empire initialization is carried out by adopting the progressive Latin hypercube sampling method, which improves the uneven distribution of the empire in the search space and expands the search range of the algorithm. This algorithm introduces imperialism splitting mechanism into the revolution process, so that it can quickly jump out of the local optimum and effectively avoid premature convergence. Besides, it introduces clonal evolution operator in competition process to reproduce the countries with strong influence. The clonal populations are randomly crossed and high-efficiency mutated to guide the algorithm forward in the optimal solution direction. Furthermore, the support–confidence framework is extended to form a rule evaluation system consisting of support, confidence, interest, comprehensible and Kulczynski metric. The rule evaluation system significantly improves the quality of rules and obtains more meaningful rules. Finally, the proposed method is used to mine the data in the historical case database of NC machine tool enterprises. The method has fast solving speed and can obtain high-quality rules.