基于马氏gm(11)和bp神经网络的货运公司收益问题学位论文(5)

SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesTSelectionParbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbagraphFoLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointse11111111111111111111111111111111lectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraphFormatLineSpaci2222222222222222222222ngLinesToPoints2SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraphFccccccccccccccccccccccccccccccccccccccccccccccccccccccccormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraSelec

如图三所示：

o：表示实际值 AND *：表示预测值60005000400030002000100000510152025303540 图三 马氏GM(1,1)模型预测值与实际值的比较

在图三中我们可以看到，经过优化，预测值相比较GM(1,1)模型更接近于实际值。

误差曲线为：

优化模型一的误差曲线15001000500相对误差值0-500-1000-1500051015日期202530 图四 马氏GM(1,1)模型的误差曲线

SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesTSelectionParbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbagraphFoLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointse11111111111111111111111111111111lectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraphFormatLineSpaci2222222222222222222222ngLinesToPoints2SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraphFccccccccccccccccccccccccccccccccccccccccccccccccccccccccormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraSelec

11

SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesTSelectionParbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbagraphFoLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointse11111111111111111111111111111111lectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraphFormatLineSpaci2222222222222222222222ngLinesToPoints2SelectionParagraphFormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPointselectionParagraphFccccccccccccccccccccccccccccccccccccccccccccccccccccccccormatLineSpacingLinesToPointsSelectionParagraphFormatLineSpacingLinesToPoctionParagraSelec

图四中可以得出，每天预测值的误差经过优化减少了很多，达到了一个可以令人接受的程度。之后我们又对F类G类H类货物申请量的情况作了优化，都得到了基本令人满意的结果。

最终，在GM (1,1)模型基础上经过马尔可夫链的优化后，我们得到了四类货物7天的申请量的预测值，如下表所示：

表五 四类货物7天申请量的预测值（kg）

E类 F类 7.1 1881 3798 6091 4716 7.2 2661 2968 4456 6219 7.3 1922 3426 4601 1961 7.4 2598 3873 4749 663 7.5 1874 2935 4901 2165 7.6 2251 3376 5056 3668 7.7 1828 3848 5214 2370 G类 H类 4.2.3.1模型二 BP神经网络模型 …… 此处隐藏：1142字，全部文档内容请下载后查看。喜欢就下载吧 ……