To solve a dynamic multi-objective optimization problem better, algorithms need to quickly adapt to environmental changes and track its changing Pareto fronts fast. In this paper, an algorithm (SPTr-RMMEDA) based on special point and transfer component analysis is presented, in which, by using some special points and their neighborhoods solutions, the prediction strategy are used to generated the transfer learning prediction model, together with part new initial solutions to generate next population when change occurs. In order to better adapt to environmental changes, in addition to reusing some important historical information by transfer component analysis, the algorithm also introduces the adaptive diversity introduction strategy. The algorithm performs on 12 groups test problems. And the experimental results show that the proposed algorithm is superior to other four algorithms, such as Tr-NSGA-II, Tr-RMMEDA, DNSGA-II-A and DNSGA-II-B on most test problems in term of convergence and computation complexity.