基于深度网络和船舶交通流的航道水深预测方法研究
|
摘要
根据水上交通的特点,提出了一种基于船舶自动识别系统(AIS)大数据,构建深度网络模型预测航道水深的方法,并利用最新航道水深数据作为标签验证.分别利用深度神经网络算法和决策树-深度神经网络结合的DT-NN算法,对水深数据和AIS数据进行学习.实验结果表明,深度神经网络算法的预测准确度为90.84%,DT-NN算法的预测准确度为91.15%,因此,采用决策树和深度神经网络结合的DT-NN算法对于水深预测的模型准确率较高,对于弥补航道水深数据的不足,指导船舶安全航行.
According to the characteristics of water transportation, this paper proposed a method to build a depth network model to predict channel water depth based on ship automatic identification system(AIS) big data, and used the latest channel water depth data as label verification. The depth neural network algorithm and DT-NN algorithm combined with decision tree and depth neural network were used to learn water depth data and AIS data respectively. The experimental results show that the prediction accuracy of the deep neural network algorithm is 90.84%, and the prediction accuracy of DT-NN algorithm is 91.15%. Therefore, the DT-NN algorithm using the combination of decision tree and deep neural network has higher accuracy for water depth prediction, and plays an important role in making up for the shortage of channel water depth data and guiding the safe navigation of ships.
According to the characteristics of water transportation, this paper proposed a method to build a depth network model to predict channel water depth based on ship automatic identification system(AIS) big data, and used the latest channel water depth data as label verification. The depth neural network algorithm and DT-NN algorithm combined with decision tree and depth neural network were used to learn water depth data and AIS data respectively. The experimental results show that the prediction accuracy of the deep neural network algorithm is 90.84%, and the prediction accuracy of DT-NN algorithm is 91.15%. Therefore, the DT-NN algorithm using the combination of decision tree and deep neural network has higher accuracy for water depth prediction, and plays an important role in making up for the shortage of channel water depth data and guiding the safe navigation of ships.
引文
[1] SU H, LIU H, WU Q. Prediction of water depth from multispectral satellite imagery—the regression kriging alternative[J]. IEEE Geoscience & Remote Sensing Letters, 2015,12(12):2511-2515.
[2] SHIRI J. Wavelet and neuro-fuzzy conjunction model for predicting water table depth fluctuations[J]. Hydrology Research, 2012,43(1):286-300.
[3] WANG Y G. Modified rayleigh distribution of wave heights in transitional water depths[J]. Chinese Ocean Engineering, 2016,30(3):447-458.
[4] YOUNIS B A, SOUSA V, MEIRELES I. Prediction of the asymptotic water depth in rough compound channels[J]. Journal of Irrigation & Drainage Engineering, 2009,135(2):231-234.
[5] HAN S, LIU X, MAO H, et al. EIE: efficient inference engine on compressed deep neural network[C].International Symposium on Computer Architecture,Shanghai, 2016.
[6] KAMIRAN F, CALDERS T, PECHENIZKIY M. Discrimination aware decision tree learning[J]. Nature,2015(2):869-874.
[7] 尚朝轩,王品,韩壮志,等.基于类决策树分类的特征层融合识别算法[J].控制与决策,2016,31(6):1009-1014.
[2] SHIRI J. Wavelet and neuro-fuzzy conjunction model for predicting water table depth fluctuations[J]. Hydrology Research, 2012,43(1):286-300.
[3] WANG Y G. Modified rayleigh distribution of wave heights in transitional water depths[J]. Chinese Ocean Engineering, 2016,30(3):447-458.
[4] YOUNIS B A, SOUSA V, MEIRELES I. Prediction of the asymptotic water depth in rough compound channels[J]. Journal of Irrigation & Drainage Engineering, 2009,135(2):231-234.
[5] HAN S, LIU X, MAO H, et al. EIE: efficient inference engine on compressed deep neural network[C].International Symposium on Computer Architecture,Shanghai, 2016.
[6] KAMIRAN F, CALDERS T, PECHENIZKIY M. Discrimination aware decision tree learning[J]. Nature,2015(2):869-874.
[7] 尚朝轩,王品,韩壮志,等.基于类决策树分类的特征层融合识别算法[J].控制与决策,2016,31(6):1009-1014.