病毒性乙型肝炎2011年度预警预测(ARIMA模型)
2011年乙型肝炎发病预警与预测(ARIMA模型)根据绩效考核工作及上级部门的要求,我中心在2011年年初对我旗乙型肝炎发病情况进行了预警与预测,为制定相应防治措施提供依据,具体预警与预测结果如下:
一、基本数据
观测值 年份 月份
18.00 2005 1 JAN 2005
12.00 2005 2 FEB 2005
25.00 2005 3 MAR 2005
13.00 2005 4 APR 2005
14.00 2005 5 MAY 2005
20.00 2005 6 JUN 2005
14.00 2005 7 JUL 2005
25.00 2005 8 AUG 2005
15.00 2005 9 SEP 2005
21.00 2005 10 OCT 2005
27.00 2005 11 NOV 2005
18.00 2005 12 DEC 2005
18.00 2006 1 JAN 2006
21.00 2006 2 FEB 2006
33.00 2006 3 MAR 2006
31.00 2006 4 APR 2006
20.00 2006 5 MAY 2006
19.00 2006 6 JUN 2006
23.00 2006 7 JUL 2006
25.00 2006 8 AUG 2006
23.00 2006 9 SEP 2006
18.00 2006 10 OCT 2006
21.00 2006 11 NOV 2006
26.00 2006 12 DEC 2006
23.00 2007 1 JAN 2007
12.00 2007 2 FEB 2007
24.00 2007 3 MAR 2007
22.00 2007 4 APR 2007
25.00 2007 5 MAY 2007
31.00 2007 6 JUN 2007
35.00 2007 7 JUL 2007
30.00 2007 8 AUG 2007
16.00 2007 9 SEP 2007
18.00 2007 10 OCT 2007
26.00 2007 11 NOV 2007
41.00 2007 12 DEC 2007
12.00 2008 1 JAN 2008
14.00 2008 2 FEB 2008
16.00 2008 3 MAR 2008
26.00 2008 4 APR 2008
30.00 2008 5 MAY 2008
14.00 2008 6 JUN 2008
22.00 2008 7 JUL 2008
24.00 2008 8 AUG 2008
25.00 2008 9 SEP 2008
26.00 2008 10 OCT 2008
19.00 2008 11 NOV 2008
11.00 2008 12 DEC 2008
10.00 2009 1 JAN 2009
22.00 2009 2 FEB 2009
28.00 2009 3 MAR 2009
21.00 2009 4 APR 2009
27.00 2009 5 MAY 2009
19.00 2009 6 JUN 2009
26.00 2009 7 JUL 2009
26.00 2009 8 AUG 2009
12.00 2009 9 SEP 2009
20.00 2009 10 OCT 2009
21.00 2009 11 NOV 2009
20.00 2009 12 DEC 2009
23.00 2010 1 JAN 2010
14.00 2010 2 FEB 2010
33.00 2010 3 MAR 2010
34.00 2010 4 APR 2010
13.00 2010 5 MAY 2010
19.00 2010 6 JUN 2010
19.00 2010 7 JUL 2010
29.00 2010 8 AUG 2010
23.00 2010 9 SEP 2010
21.00 2010 10 OCT 2010
26.00 2010 11 NOV 2010
29.00 2010 12 DEC 2010
二、数据处理
1、时间序列
根据时间序列要求,首先做观测值与时间的序列图,如图1,序列图整体印象为年发病趋势较为平稳,没有发现季节性趋势。
2、去除趋势
(1)去除季节性差分乙型肝炎发病趋势的整体趋势如图2。
三、ARIMA(p,d,q)*(sp,sd,sq)模型建立
1、模型识别
主要通过解读自相关(ACF)和偏相关(PACF)把握模型的大致方向,为目标序列定阶,提供几个粗模型以便进一步分析完善。
(1)自相关图
从ACF图中可以看出所有自相关系数基本落入置信区间以内,说明该时间序列具有较好的平稳性。
(2)偏相关图
图中显示该时间序列具有较好的平稳性,可以用于建模。
通过观察ACF和PACF图形,可以确定对原始数据进行一次差分,数据较平稳,可以初步确定d=1,sd=0,并且两图中自相关系数判定模型阶数(p,q)的取值,采用从低阶向高阶不断尝试的办法,通过比较各个模型的AIC和SBC值(去两值最小的模型),sp、sq一般情况取0、1,较少去到2的情况。
根据取小原则,我们选择ARIMA(p=0,d=1,q=0)*(sp=0,sd=1,sq=0)模型为最适合模型。
四、疾病预测
(1)预测结果
月份 预测值 标准误 上限 下限
FEB 2005 19.67850 -7.67850 3.76964 35.58736 7.97249
MAR 2005 18.46191 6.53809 2.65446 34.26937 7.92167
APR 2005 22.68437 -9.68437 6.95426 38.41449 7.88291
MAY 2005 19.75019 -5.75019 4.07301 35.42738 7.85639
JUN 2005 20.53747 -.53747 4.88855 36.18639 7.84223
JUL 2005 22.75607 -8.75607 7.11062 38.40152 7.84049
AUG 2005 21.53948 3.46052 5.87268 37.20629 7.85119
SEP 2005 25.18942 -10.18942 9.47655 40.90228 7.87427
OCT 2005 22.82777 -1.82777 7.04435 38.61119 7.90963
NOV 2005 25.04637 1.95363 9.16822 40.92452 7.95710
DEC 2005 27.26497 -9.26497 11.26836 43.26159 8.01647
JAN 2006 16.76615 1.23385 .73228 32.80001 8.03514
FEB 2006 19.67850 1.32150 3.76964 35.58736 7.97249
MAR 2006 21.03830 11.96170 5.23085 36.84576 7.92167
APR 2006 24.97450 6.02550 9.24439 40.70461 7.88291
MAY 2006 24.90298 -4.90298 9.22579 40.58016 7.85639
JUN 2006 22.25506 -3.25506 6.60614 37.90398 7.84223
JUL 2006 22.46981 .53019 6.82435 38.11526 7.84049
AUG 2006 24.11588 .88412 8.44907 39.78268 7.85119
SEP 2006 25.18942 -2.18942 9.47655 40.90228 7.87427
OCT 2006 25.11789 -7.11789 9.33447 40.90131 7.90963
NOV 2006 24.18757 -3.18757 8.30942 40.06572 7.95710
DEC 2006 25.54738 .45262 9.55076 41.54399 8.01647
JAN 2007 19.05627 3.94373 3.02241 35.09014 8.03514
FEB 2007 21.10983 -9.10983 5.20097 37.01868 7.97249
MAR 2007 18.46191 5.53809 2.65446 34.26937 7.92167
APR 2007 22.39811 -.39811 6.66800 38.12822 7.88291
MAY 2007 22.32659 2.67341 6.64940 38.00377 7.85639
JUN 2007 23.68639 7.31361 8.03747 39.33531 7.84223
JUL 2007 25.90499 9.09501 10.25954 41.55045 7.84049
AUG 2007 27.55107 2.44893 11.88426 43.21787 7.85119
SEP 2007 26.62074 -10.62074 10.90788 42.33361 7.87427
OCT 2007 23.11403 -5.11403 7.33061 38.89745 7.90963
NOV 2007 24.18757 1.81243 8.30942 40.06572 7.95710
DEC 2007 26.97871 14.02129 10.98209 42.97532 8.01647
JAN 2008 23.35026 -11.35026 7.31640 39.38412 8.03514
FEB 2008 17.96090 -3.96090 2.05205 33.86976 7.97249
MAR 2008 19.03444 -3.03444 3.22699 34.84190 7.92167
APR 2008 20.10798 5.89202 4.37787 35.83809 7.88291
MAY 2008 23.47165 6.52835 7.79447 39.14883 7.85639
JUN 2008 25.11772 -11.11772 9.46880 40.76664 7.84223
JUL 2008 21.03848 .96152 5.39302 36.68393 7.84049
AUG 2008 23.82961 .17039 8.16281 39.49641 7.85119
SEP 2008 24.90315 .09685 9.19029 40.61601 7.87427
OCT 2008 25.69042 .30958 9.90700 41.47384 7.90963
NOV 2008 26.47770 -7.47770 10.59955 42.35585 7.95710
DEC 2008 24.97485 -13.97485 8.97823 40.97146 8.01647
JAN 2009 14.76229 -4.76229 -1.27158 30.79615 8.03514
FEB 2009 17.38837 4.61163 1.47952 33.29723 7.97249
MAR 2009 21.32457 6.67543 5.51712 37.13202 7.92167
APR 2009 23.54317 -2.54317 7.81306 39.27328 7.88291
MAY 2009 22.04032 4.95968 6.36314 37.71750 7.85639
JUN 2009 24.25892 -5.25892 8.61000 39.90784 7.84223
JUL 2009 22.46981 3.53019 6.82435 38.11526 7.84049
AUG 2009 24.97467 1.02533 9.30787 40.64148 7.85119
SEP 2009 25.47568 -13.47568 9.76282 41.18854 7.87427
OCT 2009 21.96897 -1.96897 6.18555 37.75239 7.90963
NOV 2009 24.76010 -3.76010 8.88196 40.63825 7.95710
DEC 2009 25.54738 -5.54738 9.55076 41.54399 8.01647
JAN 2010 17.33868 . 1.30482 33.37254 8.03514
FEB 2010 19.48918 . 2.97836 36.00001 8.27416
MAR 2010 20.60581 . 4.14402 37.06760 8.24958
APR 2010 21.42647 . 5.03494 37.81800 8.21437
MAY 2010 22.16240 . 5.82133 38.50348 8.18909
JUN 2010 22.87408 . 6.56010 39.18807 8.17552
JUL 2010 23.57882 . 7.26816 39.88949 8.17385
AUG 2010 24.28157 . 7.95043 40.61272 8.18411
SEP 2010 24.98375 . 8.60842 41.35909 8.20626
OCT 2010 25.68577 . 9.24272 42.12882 8.24019
NOV 2010 26.38775 . 9.85375 42.92175 8.28577
DEC 2010 27.08970 . 10.44190 43.73751 8.34280
模型拟合情况见下图,预测值与实际值基本吻合,表明模型选择是正确的,拟合效果较好。
(2)结果验证
对生成数据的残差做自相关和偏相关分析,结果残差是白噪声,各个参数有意义。
五、意见反馈
根据预测,2011年我旗乙型肝炎发病(282例)较2011年(283例)相对持平,乙型肝炎一直是我旗发病率占前三位的传染病,是我旗传染病防控重点,加强医疗机构对乙型肝炎的管理,防止医源性感染,另一方面应加强乙型肝炎防治知识的宣传力度,提高人群对乙型肝炎的认识,做到防病治病,有效控制乙型肝炎的发生。 来支持一下,做得很好 强啊,我也想尝试一下
页:
[1]