User's Preference Aggregation Based on Parallel Interactive Genetic Algorithms

摘要

In traditional interactive genetic algorithms, high-quality optimal solution is hard to be obtained due to small population size and limited evolutional generations. Aming at above problems, a parallel interactive genetic algorithm based on knowledge migration is proposed. During thee volution, the number of the populations is more than one. Evolution information can be exchanged between every two populations so as to guide themselves evolution. In order to realize the freedom communication, IP multicast is adopted as the transfer protocol to find out the similar users instead of traditional TCP/IP communication mode. Taken the fashion evolutionary design system as test platform, the results indicate that the IP multicast-based parallel interactive genetic algorithm has better population diversity. It also can alleviate user fatigue and speed up the convergence.