forecast :: thetaf错误

时间:2018-03-02 10:15:54

标签: r time-series forecasting

我目前在预测包8.2和thetaf函数方面遇到问题:

selector = {'scope_id': {'$all': [tool_id, api_id, input_params]}}

update = {
    'scope_id': [
        tool_id, api_id, input_params, 
        datetime.now().strftime('%Y-%m-%d %H:%M:%S')
     ]
}
users.update(selector, update, upsert=True)

给出以下错误

dat<- structure(c(5, 0, 5, 0, 0, 2, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.9, 0, 2, 0, 1, 
0, 2.1, 0, 2, 0, 1, 0, 0, 0, 2.5, 2, 2, 0, 1.7, 0, 1.5, 0, 1, 
0, 0, 0, 2.5, 0), .Tsp = c(1999, 2003.91666666667, 12), class = "ts")
library(forecast)
thetaf(dat,h = 1)$mean

错误与频率有关。如果它是1而不是12然后它工作。我没有看到为什么这会导致错误的原因。有任何想法吗?

1 个答案:

答案 0 :(得分:1)

当应用于季节性数据时,theta方法使用乘法经典分解来消除季节性。在这种情况下,分解失败:

> decompose(dat, type="multiplicative")
$x
     Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1999 5.0 0.0 5.0 0.0 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0
2000 4.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
2001 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.9 0.0 2.0 0.0
2002 1.0 0.0 2.1 0.0 2.0 0.0 1.0 0.0 0.0 0.0 2.5 2.0
2003 2.0 0.0 1.7 0.0 1.5 0.0 1.0 0.0 0.0 0.0 2.5 0.0

$seasonal
           Jan       Feb       Mar       Apr       May       Jun       Jul       Aug
1999 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2000 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2001 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2002 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2003 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
           Sep       Oct       Nov       Dec
1999 0.7674245 0.0000000 2.1696136 0.7889410
2000 0.7674245 0.0000000 2.1696136 0.7889410
2001 0.7674245 0.0000000 2.1696136 0.7889410
2002 0.7674245 0.0000000 2.1696136 0.7889410
2003 0.7674245 0.0000000 2.1696136 0.7889410

$trend
           Jan       Feb       Mar       Apr       May       Jun       Jul       Aug
1999        NA        NA        NA        NA        NA        NA 0.9583333 0.9166667
2000 0.3333333 0.3333333 0.3333333 0.3333333 0.3333333 0.3333333 0.1666667 0.0000000
2001 0.0000000 0.0000000 0.0375000 0.0750000 0.1583333 0.2416667 0.2833333 0.3250000
2002 0.7083333 0.7500000 0.7125000 0.6750000 0.6958333 0.8000000 0.9250000 0.9666667
2003 0.8916667 0.8916667 0.8916667 0.8916667 0.8916667 0.8083333        NA        NA
           Sep       Oct       Nov       Dec
1999 0.7083333 0.5000000 0.5000000 0.4166667
2000 0.0000000 0.0000000 0.0000000 0.0000000
2001 0.4125000 0.5000000 0.5833333 0.6666667
2002 0.9500000 0.9333333 0.9125000 0.8916667
2003        NA        NA        NA        NA

$random
           Jan       Feb       Mar       Apr       May       Jun       Jul       Aug
1999        NA       NaN        NA       NaN        NA        NA 0.0000000       NaN
2000 2.1793043       NaN 0.0000000       NaN 0.0000000       NaN 0.0000000       NaN
2001       NaN       NaN 0.0000000       NaN 0.0000000       NaN 0.0000000       NaN
2002 0.2563887       NaN 2.3017827       NaN 2.3911982       NaN 3.7907196       NaN
2003 0.4073466       NaN 1.4889369       NaN 1.3995214       NaN        NA       NaN
           Sep       Oct       Nov       Dec
1999 0.0000000       NaN 0.0000000 0.0000000
2000       NaN       NaN       NaN       NaN
2001 2.8430397       NaN 1.5802682 0.0000000
2002 0.0000000       NaN 1.2627714 2.8430397
2003        NA       NaN        NA        NA

$figure
 [1] 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
 [9] 0.7674245 0.0000000 2.1696136 0.7889410

$type
[1] "multiplicative"

attr(,"class")
[1] "decomposed.ts"

问题出现是因为四月,六月,八月和十月的季节性指数为零,所以除以季节性指数就会产生NaN值。

我已经解决了问题(在预测包的v8.3中),以便在发生这种情况时继续使用theta方法的非季节性版本。