Study on prediction of ship’s power using light GBM and XGBoost
0
조회
0
다운로드
4
인용
0
참고문헌
키워드
초록
Research on smart ship is becoming active. In smart ships, forecasting power demand is important. Through the power forecast, the power supply can be flexibly improved to maximize the power generation efficiency. Energy reduction can be achieved through increased efficiency. In this paper, we construct a ship power demand prediction model using XGBoost (eXtream Gradient Boost) and LGBM (Light Gradient Boosting Model). The hyperparameters of the boosting algorithm improve the accuracy of the prediction model and prevent overfitting. In addition, the verification functions of XGBoost and LGBM were constructed and verified, and the model suitability was compared. The verification function used Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Absolute Scaled Error (MASE). The XGBoost model showed similar performance compared to LGBM, but was slow in computation speed. In terms of operation speed, LGBM is recommended when constructing a power demand prediction model using boosting techniques in smart ships.
Abstract
Research on smart ship is becoming active. In smart ships, forecasting power demand is important. Through the power forecast, the power supply can be flexibly improved to maximize the power generation efficiency. Energy reduction can be achieved through increased efficiency. In this paper, we construct a ship power demand prediction model using XGBoost (eXtream Gradient Boost) and LGBM (Light Gradient Boosting Model). The hyperparameters of the boosting algorithm improve the accuracy of the prediction model and prevent overfitting. In addition, the verification functions of XGBoost and LGBM were constructed and verified, and the model suitability was compared. The verification function used Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Absolute Scaled Error (MASE). The XGBoost model showed similar performance compared to LGBM, but was slow in computation speed. In terms of operation speed, LGBM is recommended when constructing a power demand prediction model using boosting techniques in smart ships.
참고문헌 (6)
- T. Chen, "XGBoost: A Scalable Tree Boosting System", Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp.785~794 (2016)
- G. Ke, "LightGBM : A highly efficient gradi-ent boosting decision tree", NIPS Proceeding of 31st Con-ference, pp.3146~3154 (2017)
- 김연근, "Data Fusion 기술을 활용한 스마트선박 내 효율적 데이터 전송 방안", 한국통신학회논문지C, 39(11), pp.1146~1150 (2014), 한국통신학회
- J. H. Park, "Trends in smart monitor-ing and failure handling systems for ships and offshore plants", The Korean Institute of Information Scientists and Engineers(in Korean), 31(1), pp.46~54 (2013)
- S. M. Jeon, "Intelligent energy management system for smart ship", Conference Proceeding (summer) of Korea Institute of Communication Sciences, pp.209~210 (2013)
- S. H. Jeong, "The common platform technology of smart maritime autonomous sur-face ships", Proceedings of KIIT Conference, pp.442~445 (2018)