CAJ | 학술논문

为提高船舶市场趋势预测的精度，针对以往在神经网络应用时仅单纯改进隐层环节算法的局限性，综合模糊聚类方法、数据修正和插值算法，对输入环节的数据进行降维和增量处理，构建船舶市场趋势预测的三阶段模型。首先，利用模糊聚类方法对历史数据进行分类，降低数据非线性；然后，通过数据修正和插值算法，在不改变数据规律的情况下增加每类数据的数据量；最后，利用处理完毕的数据训练神经网络。实例结果证明，三阶段模型在船舶市场趋势预测方面是有效的。
위제고선박시장추세예측적정도，침대이왕재신경망락응용시부단순개진은층배절산법적국한성，종합모호취류방법、수거수정화삽치산법，대수입배절적수거진행강유화증량처리，구건선박시장추세예측적삼계단모형。수선，이용모호취류방법대역사수거진행분류，강저수거비선성；연후，통과수거수정화삽치산법，재불개변수거규률적정황하증가매류수거적수거량；최후，이용처리완필적수거훈련신경망락。실례결과증명，삼계단모형재선박시장추세예측방면시유효적。
In order to improve the prediction accuracy of ship market trend,in view of the limitation that the neural network is simply used to improve the algorithm in the hidden layer,the reduction of data di-mension and the increase of the data amount are carried out in the input layer by the fuzzy clustering method,data correction and interpolation algorithm,and a three-stage model of ship market trend predic-tion is constructed. First,the fuzzy clustering method is used to classify the historical data to reduce the nonlinearity of data. Then,through the data correction and interpolation algorithm,the amounts of vari-ous types of data increase without changing data regularity. Finally,the neural network is trained by the processed data. The application results show that the three-stage model is effective in the ship market trend prediction.