运筹学与供应链管理-第2讲ppt34.ppt

3-1McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting第讲需求预测 3-2McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting预测:是关于未来的描述预测用于帮助管理人员系统计划：产品服务的类型、设施配置、厂址选择实施计划：库存、劳动力、采购、生产进度 3-3McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting会计成本/利润估计财务现金流量人力资源招工/招聘/培训营销价格,促销,战略管理信息系统服务运作调度,MRP,工作负荷产品/服务设计新产品和服务预测的用途 3-4McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting假定系统中存在因果关系过去=未来由于不确定性因素的存在使得预测很少准确无误对一组事物的预测比 对单个事物预测准确当预测时间跨度增加时，预测精度将下降 我看这学期你能得优秀成绩。3-5McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting好的预测方法的基本要素 适时适时 精度精度 可靠可靠 含义含义 书面书面 易用易用 3-6McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting预测进行的步骤步骤步骤 1 确定预测目的确定预测目的步骤步骤2 确定预测时间跨度确定预测时间跨度步骤步骤3 选择预测技术选择预测技术步骤步骤4 收集和分析数据收集和分析数据步骤步骤 5 准备预测准备预测步骤步骤 6 预测监控预测监控 “预测预测”3-7McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting预测的类型主观判断 主观意见时间序列 使用历史数据，认为将来和过去相似联合模型 基于自变量预测未来 3-8McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting主观预测主观预测经理人员的意见与顾客直接接触人员的意见消费者调查其它预测方法德尔非法Delphi method 3-9McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting时间序列预测长期趋势 数据的长期移动季节性变动 数据短期规则变化不规则变动 异常情况引起的变动随机变动 各种可能性引起的变动 3-10McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-7 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e需求变动的形式需求时间趋势变动随机变动需求时间季节变动需求时间需求时间周期变动带季节性的趋势变动 3-11McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting简单易用基本上不需要成本不需要数据分析容易理解不能提供较高的精度能够作为衡量精度的标准简单预测法 3-12McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting平稳时间序列数据F(t)=A(t-1)季节变动F(t)=A(t-n)长期趋势数据F(t)=A(t-1)+(A(t-1)A(t-2)简单预测法应用 3-13McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting简单预测简单预测唉唉,给我点时间给我点时间.上周我们卖了上周我们卖了250轮胎轮胎.那么那么,下周我们应该下周我们应该卖卖.3-14McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting平均方法移动平均加权移动平均指数平滑法 3-15McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting简单移动平均图 3-4MAn=nAii=1n实际数据实际数据MA3MA5 3-16McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-14 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e加权移动平均可调整和反映简单移动平均法中不同时期数据的影响WMAn=i=1Wi Di此处,Wi=第 i 期的权重，其百分值在 0 100 之间 Wi=1.00 3-17McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-15 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e加权移动平均之例月份 权重 数据八月 17%130九月 33%110十月 50%90十一月的预测 3-18McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting指数平滑法n假设最近的观测具有最高的预测价值.因此在预测时，应该给更近的数据赋予更高的权重.Ft=Ft-1+(At-1-Ft-1)3-19McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting指数平滑法举例 3-20McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting选择平滑常数选择平滑常数 .1.4实际数据实际数据 3-21McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting线性趋势方程b 直线斜率Yt=a+bt0 1 2 3 4 5 tYb=n(ty)-tynt2-(t)2a=y-btn 3-22McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting线性趋势方程举例 3-23McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting线性趋势计算y=143.5+6.3t a=812-6.3(15)5=b=5(2499)-15(812)5(55)-225=12495-12180275-225=6.3143.5 3-24McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-25 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e季节性调节需求重复性地增加减少使用季节因子来调节预测 3-25McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-26 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e季节性调节之例需求(1000件季度)年 1 2 3 4 总计1995 12.6 8.66.317.545.01996 14.1 10.37.518.250.11997 15.3 10.68.119.653.6总计 42.0 29.5 21.955.3148.7Si 0.28 0.20 0.15 0.37趋势直线 y=40.97+4.30 x1998(第4年)的预测=40.97+4.30(4)=58.17199816.28 11.63 8.73 21.53SF1=(S1)(F5)=(0.28)(58.17)=16.28SF2=(S2)(F5)=(0.20)(58.17)=11.63SF3=(S3)(F5)=(0.15)(58.17)=8.73SF4=(S4)(F5)=(0.37)(58.17)=21.53 3-26McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting关联预测有关的变量 预示这些量的值回归分析 用一条线去拟合一组点集最小二乘法 使线与点间的距离平方和最小 3-27McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting线性模型经计算得出的因果关系 3-28McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-37 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e回归方法研究两个或更多变量之间的关系因变量随自变量而变 3-29McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-38 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e线性回归计算y=a+bx 此处,a=截距b=直线斜率x=自变量y=给定 x 时的预测值b=a=y-b x此处,n=期数x=x ,x 的平均值 ny=y ,y 的平均值 n x2-nx2xy-nxy 3-30McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-39 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e线性回归之例xyxyx2436.3145.216640.1240.636641.2247.236853.042464644.026436745.6319.249539.019525747.5332.54949346.72167.7311 3-31McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingCh 10-40 2000 by Prentice-Hall IncRussell/Taylor Oper Mgt 3/e线性回归直线 3-32McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting预测精度误差 实际值与预测值之间的偏差绝对平均偏差(MAD)误差的绝对值的平均值均方差(MSE)误差的平方的平均值跟踪信号累计误差与 MAD 之间的比率 3-33McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.ForecastingMAD&MSEMAD=实际值 预测值nMSE=实际值 预测值)-12n(3-34McGraw-Hill/IrwinOperations Management,Seventh Edition,by William J.StevensonCopyright 2002 by The McGraw-Hill Companies,Inc.All rights reserved.Forecasting跟踪信号 跟踪信号=(实际值 -预测值)MAD 跟踪信号=(实际值 -预测值)实际值 -预测值n