상세정보
학술대회
Objectivity based Self-Evolving Agent Model Validation for Social Issue Simulation
- 발행일
- 201810
- 출처
- International Conference on Information and Communication Technology Convergence (ICTC) 2018, pp.1417-1419
- 협약과제
- 18HS3200, 점진적 기계학습 기반 자가진화(Self-Evolving) 에이전트 시뮬레이션을 이용한 사회변화 예측분석 기술 개발, 백의현
- 초록
- In this paper, objectivity based self-evolving agent model validation is presented for house market simulation. The self-evolving simulation updates agent models using verification data. However, because there is no verification data for the future, it is difficult to predict the future with self-evolving module accurately. Therefore, even if simulation is performed using a self-evolving mechanism, validation of the agent model is still important in order to obtain correct prediction results for the future. Verification of objectivity is carried out in the procedure of comparing the simulation prediction value with the past actual data, and it is judged that the result of the simulation is valid if it shows a form similar to the past actual data. The error rate and correlation between the verification data and simulation results are calculated through the root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination. The accuracy of the simulation results for the apartment price index and 2-year-lease index are approximately 95% and 89%, respectively, and coefficient of determination are 0.507 and 0.753, respectively. Therefore, self-evolving agent based model designed for house market simulation can be said to have objectivity.
- KSP 제안 키워드
- Agent model, Coefficient of determination, Market simulation, Model Validation, Price index, Root mean square(RMS), Simulation prediction, Social issues, Verification data, agent-based model(ABM), error rate