본 연구는 주식 포트폴리오 전략으로 유전자알고리즘을 사용한 펀드 표준화 모형을 적용하고 슬라이딩 윈도우 기법을 통해서 투자 최적화 방법을 찾기 위해서 기간별 투자성과를 분석했다. 실증분석 주요 결과는 다음과 같다. 우선, 유전자알고리즘에 샤프 비율을 도입하여 투자 최적화를 도입한 경우는 코스피200보다 높은 샤프 비율을 보여주었다. 학습 기간이 길수록 샤프 비율이 높은 최적의 조합을 찾았다. 두 번째는 슬라이딩 윈도우를 활용한 최적 펀드 투자모델 분석에서는 장기 데이터를 기반으로 보았을 때는 코스피200보다는 모든 투자모델이 매우 웃도는 투자실적을 나타내고 있다. 특히 매매 기간이 2주 단위의 경우가 월 단위 매매보다 좋은 성과를 보였다. 반면에 주가 변동성이 큰 구간을 제외한 단기 데이터를 바탕에서는 모델 간 차이가 존재했다. 따라서 학습-매매 기간의 변화를 충분히 반영해야 한다. 마지막으로 손절매 전략은 유용한 것으로 나타났다. -5%까지는 투자수익률을 방어하지만, 그 이하는 투자수익률이 낮아지는 것을 확인했다.
This study aims to analyze investment performance by dividing the investment optimization method by period through a sliding window method and fund standardization using genetic algorithm as a portfolio optimization method, The research results of the empirical analysis are as follows. First, the investment optimization using the genetic algorithm showed a higher Sharpe ratio than the KOSPI 200. The longer the learning period, the higher the Sharpe ratio was found. Second, the optimal fund investment model using a sliding window, based on long-term period, all investment models show much higher investment performance than the KOSPI 200. In particular, two-week periods of trading performed better than monthly trading. On the other hand, there were differences between models based on short-term data excluding the period with high return volatility. Therefore, it is necessary to sufficiently reflect the change in the learning-testing period. Finally, the loss-cut strategy was found to be useful. It was confirmed that the ROI was defended up to -5%, but the lower ROI was rather worse than that.
This study aims to analyze investment performance by dividing the investment optimization method by period through a sliding window method and fund standardization using genetic algorithm as a portfolio optimization method, The research results of the empirical analysis are as follows. First, the investment optimization using the genetic algorithm showed a higher Sharpe ratio than the KOSPI 200. The longer the learning period, the higher the Sharpe ratio was found. Second, the optimal fund investment model using a sliding window, based on long-term period, all investment models show much higher investment performance than the KOSPI 200. In particular, two-week periods of trading performed better than monthly trading. On the other hand, there were differences between models based on short-term data excluding the period with high return volatility. Therefore, it is necessary to sufficiently reflect the change in the learning-testing period. Finally, the loss-cut strategy was found to be useful. It was confirmed that the ROI was defended up to -5%, but the lower ROI was rather worse than that.
