Product redesign is widely required in the process of product innovation, which greatly depends on a series of innovative theories of product. Traditional product redesign can identify functional requirements from customer needs by functional analysis and utilize inventive principles of TRIZ to address the problems. However, the relationships between the customer needs (CNs), engineering parameters (EPs), and inventive principles (IPs) has not been explored, which may limit and hinder the opportunity for product innovation. To improve the design efficiency, Random Forest (RF) is introduced in this method to mine the relationship between EPs and IPs by researching many cases for design knowledge. A systematic redesign process is proposed to apply the design knowledge. Firstly, functional analysis is used to decompose customer needs (CNs) into product functions. And then, the effect-based approach which is the application of scientific laws is used to transform functions into EPs. Secondly, IPs are recommended based on EPs by a trained RF model. Finally, the specific solution is generated from the problem domain using the analogy design based on the Su-Field model. The effectiveness of the method is verified in the redesign of solar power systems for unmanned transport ships.