Parallel Problem Solving from Nature - PPSN XII

Volume 7492 of the series Lecture Notes in Computer Science pp 498-507

Finding Good Affinity Patterns for Matchmaking Parties Assignment through Evolutionary Computation

  • Sho KuroiwaAffiliated withNara Institute of Science and TechnologyHopeful Monster Corporation
  • , Keiichi YasumotoAffiliated withNara Institute of Science and Technology
  • , Yoshihiro MurataAffiliated withHiroshima City University
  • , Minoru ItoAffiliated withNara Institute of Science and Technology

* Final gross prices may vary according to local VAT.

Get Access


There is a demand to maximize the number of successful couples in matchmaking parties called “Gokon” in Japanese. In this paper, we propose a method to find good affinity patterns between men and women from resulting Gokon matches by encoding their attribute information into solutions and using an evolutionary computation scheme. We also propose a system to assign the best members to Gokons based on the method. To derive good affinity patterns, a specified number of solutions as chromosomes of evolutionary computation (EC) are initially prepared in the system. By feeding back the results of Gokon to the solutions as fitness value of EC, semi-optimal solutions are derived. To realize the proposed system, we need simultaneous search of multiple different good affinity patterns and efficient evaluation of solutions through as small number of Gokons as possible with various attribute members. To meet these challenges, we devise new methods for efficient selection operation inspired by Multi-niches Crowding method and reuse of past Gokon results to evaluate new solutions. To evaluate the system, we used the NMax problem assuming that there would be N good affinity patterns between men and women as a benchmark test. Through computer simulations for N = 12, we confirmed that the proposed system achieves almost twice as many good matches as a conventional method with about half the evaluation times.


Evolutionary Computation Matchmaking Party Multi- niches Crowding