Adjusted k-nearest neighbor algorithm: Application on forecasting the ship accident in South Korea
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
In this paper, we propose a new forecasting approach that finds a nearest-neighbor optimum for forecasting. It blends the advantages of subsample use (k-nearest neighbor, rolling window, etc.) and entire sample use. Basically, it belongs to the range of nearest neighbor methods but is different from them in that it also considers the entire sample, which can help to yield less variance in test error than other subsample methods. To improve forecasting accuracy, we also used a modified least squares method in the process of this forecasting approach. From simulation tests, we were able to verify that this approach yielded less test error than other methods that do not adopt this approach. Empirical verification using time-series data on the ship accident also supported that this approach is able to improve forecasting accuracy. In addition, we were able to verify that the test errors obtained from this approach were less than the residuals obtained from fitting using the actual future value in many cases. A new parameter controlling a weight between the subsample and the entire sample is introduced, and the forecasting performance may depend on how this parameter can be efficiently used.
Abstract
In this paper, we propose a new forecasting approach that finds a nearest-neighbor optimum for forecasting. It blends the advantages of subsample use (k-nearest neighbor, rolling window, etc.) and entire sample use. Basically, it belongs to the range of nearest neighbor methods but is different from them in that it also considers the entire sample, which can help to yield less variance in test error than other subsample methods. To improve forecasting accuracy, we also used a modified least squares method in the process of this forecasting approach. From simulation tests, we were able to verify that this approach yielded less test error than other methods that do not adopt this approach. Empirical verification using time-series data on the ship accident also supported that this approach is able to improve forecasting accuracy. In addition, we were able to verify that the test errors obtained from this approach were less than the residuals obtained from fitting using the actual future value in many cases. A new parameter controlling a weight between the subsample and the entire sample is introduced, and the forecasting performance may depend on how this parameter can be efficiently used.
참고문헌 (23)
- A. Inoue, "Rolling window selection for out-of-sample forecasting with time-varying parameters", Journal of Econometrics, 196(1), pp.55~67 (2017)[DOI]
- T. Ouyang, "Model of selecting prediction window in ramps forecasting", Renewable Energy, 108, pp.98~107 (2017)[DOI]
- B. Rossi, "Out-of-sample forecast tests robust to the choice of window size", Journal of Business & Economic Statistics, 30(3), pp.432~453 (2012)[DOI]
- T. Yang, "A novel prototype reduction method for the K-nearest neighbor algorithm with <TEX>$K{\geq}1$<", TEX>, 2010, 89~100
- J. W. Kim, "Nonparametric forecasting with one-sided kernel adopting pseudo one step ahead data", Proceedings of the Autumn Conference of the Korean Data and Information Science Society, pp.9~ (2010)
- 김정우, "Forecasting greenhouse gas emissions in the Korean shipping industry using the least squares adjusted with pseudo data", 한국마린엔지니어링학회지, 41(5), pp.452~460 (2017), 한국마린엔지니어링학회[DOI]
- T. Hastie, "The Elements of Statistical Learning" (2009), Springer series in statistics
- L. Zhang, "An improved K-nearest neighbor model for short-term traffic flow prediction", Procedia - Social and Behavioral Sciences, 96, pp.653~662 (2013)[DOI]
- R. Li, "Application of support vector machine combined with K-nearest neighbors in solar flare and solar proton events forecasting", Advances in Space Research, 42(9), pp.1469~1474 (2008)[DOI]
- A. T. Lora, "Time-series prediction: Application to the short-term electric energy demand", Current Topics in Artificial Intelligence Lecture Notes in Computer Science, pp.577~586 (2004)
- D. Yankov, "Ensembles of nearest neighbor forecasts", Lecture Notes in Computer Science Machine Learning: ECML 2006, pp.545~556 (2006)
- M. Pilatowska, "Information and prediction criteria in selecting the forecasting model", Dynamic Econometric Models, 11, pp.21~40 (2011)
- Y. C. Chen, "Forecasting inflation using commodity price aggregates", Journal of Econometrics, 183(1), pp.117~134 (2014)[DOI]
- M. H. Pesaran, "Selection of estimation window in the presence of breaks", Journal of Econometrics, 137(1), pp.134~161 (2007)[DOI]
- R. M. Kil, "Optimum window size for time series prediction", Proceedings of 19th Annual International Conference of the IEEE engineering in medicine and biology magazine, pp.1421~1424 (1997)
- J. M. Bates, "The combination of forecasts", Operations Research, 20(4), pp.451~468 (1969)[DOI]
- G. Elliott, "Handbook of Economic Forecasting" (2006), Elsevier
- The Shrinkage Approach in the Combination of Forecasts, Technical Report 44, "T. Wenzel", University of Dortmund, ~ (2000)
- T. E. Clark, "Improving forecast accuracy by combining recursive and rolling forecasts", International Economic Review, 50(2), pp.363~395 (2009)[DOI]
- A. Sasu, "K-nearest neighbor algorithm for univariate time series prediction", Bulletin Of The Transilvania University Of Brasov Series. III, 5(54), pp.147~152 (2012)
- 김동진, "국내 해양선박사고의 인적 오류의 요인 평가", 대한인간공학회지, 30(1), pp.87~98 (2011), 대한인간공학회
- 김홍태, "해양사고조사를 위한 인적 오류 분석사례", 대한인간공학회지, 30(1), pp.137~150 (2011), 대한인간공학회
- C. Chauvin, "Human and organisational factors in maritime accidents: Analysis of collisions at sea using the HFACS", Accident Analysis & Prevention, 59, pp.26~37 (2013)[DOI]