我实际上正在学习R上的时间序列，但是我无法将预测值转换为实际值。我为我的英语道歉，我不是本地人。我在解释自己：
这些是我的价值观：
> nbats
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2004 20 22 18 23 25 27 15 12 11 17 26 22
2005 21 25 21 25 26 29 14 12 11 17 22 22
2006 21 25 20 22 28 27 14 10 11 15 21 24
2007 22 24 20 26 29 24 14 11 10 15 23 22
2008 21 27 27 29 35 30 14 12 11 18 25 26
2009 24 28 24 29 38 29 15 11 11 18 23 24
2010 23 26 20 26 31 33 20 13 12 22 24 26
2011 25 32 26 35 42 36 11 8 8 11 12 27
2012 30 38 33 33 44 49 21 14 11 19 28 29
2013 30 34 32 36 42 51 19 11 10 19 28 29
2014 29 31 27 38 47 38 20 12 10 21 30 33
2015 30 31 27 40 47 50 17 10 11 19 33 37
2016 39 41 38 53 70 73 24 12 14 25 43 71
2017 44 48 44 68 65 69 25 15 17 39 50 51
2018 52 54 50 81 100 58 25 14 15 41 58 60
https://image.noelshack.com/fichiers/2019/17/1/1555894195-rplot01.jpeg
首先，我将数据分为两部分，以将预测值与实际值进行比较。
donnees_app = ts(donnees[1:(length(donnees)-6)],start=2004,frequency=12)
donnees_valid = ts(donnees[-(1:(length(donnees)-6))],start=c(2018,7),frequency = 12)
我的数据不是平稳的，所以我只对数据应用对数函数以对方差进行归类，然后两次相差以均值进行归类。
nba.statioVar=log(donnees_app)
nba.statioMoy=diff(nba.statioVar,lag=1)
nba.statioMoy2=diff(nba.statioVar,differences=2)
> nba.statioMoy2
Jan Feb Mar Apr May Jun Jul Aug
2004 -0.2959808753 0.4457931535 -0.1617408491 -0.0064205678 -0.6647477060 0.3646431136
2005 0.1205340690 0.2208734028 -0.3487067743 0.3487067743 -0.1351326740 0.0699785788 -0.8374377923 0.5740878205
2006 -0.0465200156 0.2208734028 -0.3974969385 0.3184537311 0.1458518770 -0.2775297010 -0.6204118922 0.3203072998
2007 -0.2205427696 0.1740227540 -0.2693329338 0.4446858213 -0.1531649725 -0.2984412916 -0.3497545011 0.2978344439
2008 -0.0020682531 0.2978344439 -0.2513144283 0.0714589640 0.1165932675 -0.3422029113 -0.6079893722 0.6079893722
2009 -0.1192634208 0.2341933875 -0.3083013597 0.3433926795 0.0810483301 -0.5405806595 -0.3889552991 0.3490907006
2010 -0.0851192288 0.1651619365 -0.3849665866 0.5247285289 -0.0864735980 -0.1133703095 -0.5632956449 0.0699923718
2011 -0.1192634208 0.2860807911 -0.4544994427 0.5048908882 -0.1149299667 -0.3364722366 -1.0314729858 0.8671699345
2012 -0.7055697006 0.1310282624 -0.3774673763 0.1410785983 0.2876820725 -0.1800514083 -0.9549285246 0.4418327523
2013 -0.0011897681 0.0912615913 -0.1857877648 0.1784076575 0.0363676442 0.0400053346 -1.1815426680 0.4408429472
2014 -0.0350913198 0.0666913745 -0.2048417130 0.4798996322 -0.1291878517 -0.4251228840 -0.4292924442 0.1310282624
2015 -0.1906203596 0.1281000026 -0.1709401613 0.5311929266 -0.2317744405 -0.0993927439 -1.1406850651 0.5481814103
2016 -0.0617666177 -0.0026333129 -0.1259963276 0.4086916608 -0.0545024253 -0.2362391294 -1.1543698099 0.4192584302
2017 -0.9799700045 0.5655016201 -0.1740227540 0.5223294482 -0.4804385065 0.1048396700 -1.0749499144 0.5044050560
2018 -0.0003845414 0.0183222421 -0.1147013691 0.5593871904 -0.2717051179 -0.7554482068
Sep Oct Nov Dec
2004 0.1361321743 0.5223294482 -0.0104348773 -0.5919372786
2005 0.0671393028 0.5223294482 -0.1774889620 -0.2578291093
2006 0.4317824164 0.2148447485 0.0263173083 -0.2029408440
2007 0.1458518770 0.5007752879 0.0219789067 -0.4718957774
2008 0.0671393028 0.5794878621 -0.1639724181 -0.2892833538
2009 0.3101549283 0.4924764851 -0.2473540271 -0.2025628436
2010 0.3507402084 0.6861785112 -0.5191244266 -0.0069686693
2011 0.3184537311 0.3184537311 -0.2314423541 0.7239188392
2012 0.1643030513 0.7877057632 -0.1587781754 -0.3526742112
2013 0.4512335266 0.7371640660 -0.2540883552 -0.3526742112
2014 0.3285040670 0.9242589015 -0.3852624008 -0.2613647641
2015 0.6259384309 0.4512335266 0.0055248759 -0.4376582311
2016 0.8472978604 0.4256678154 -0.0374942044 -0.0408445295
2017 0.6359887667 0.7051851591 -0.5818869428 -0.2286587320
2018
我使用auto.arima函数确定ARIMA模型，并保留ARIMA（3,0,3）。
所以我做了6个月的预测：
prev5 = forecast(mod5,h=6) #mod5 is the ARIMA(3,0,3) model
“ https://image.noelshack.com/fichiers/2019/17/1/1555894041-rplot.jpeg”
我希望在与第一个相同的图形中显示预测（在y轴上，值从20到100，而不是-1到1）
我不知道我是否很清楚，但我希望你能帮助我。
谢谢！
