Designing simulated annealing and evolutionary algorithm for estimating attributes of residents from statistics

In designing an agent-based social simulation model for practical problems in a real society, it is essential to estimate attributes of residents such as age, education, occupation, income, and so on. Governments often restrict a second use of personal data for protecting privacy of residents. In order to prepare personal data of residents for social simulation, some methods are developed to reconstruct personal data artificially. We have developed a simulated annealing that reconstructs attributes from available statistics in our previous work. In this paper, we develop a hybrid evolutionary algorithm to estimate citizens´ attributes from statistics. Firstly we modify a simulated annealing we have developed in order to minimize the error between the statistics of reconstructed population and the real statistics. Then we develop an evolutionary algorithm for reconstruct personal data. We apply a simulated annealing to the best solution obtained by the evolutionary algorithm to improve a solution. Our simulation result shows that the proposed hybrid algorithm gave the better results than the modified simulated annealing we have developed.