Get the prediction of test set

Jiaixang Li

2019-09-28

Assume we have

1. test dataset test
2. model xgb_base which is a xgboost model
3. we want to save it in the directory data with name my_pred.csv

phv_pred(test,xgb_base,dir='data',output_name='my_pred.csv')

---

The official target metric on the site is not usual, thus here is the function I wrap the metric into.

Assume you finish your model and get four coloumn in dataset dataset,

1. id is the id of the phv machine
2. t is time to record every y and x variables
3. p is the real power, in this contest, it is the target we want to predict.
4. phat is the predicted power, we want it to approach the real one.

library(add2evaluation)
data(dataset)
library(lubridate)
#> 
#> 载入程辑包：'lubridate'
#> The following object is masked from 'package:base':
#> 
#>     date
library(psych)
library(tidyverse)
#> ─ Attaching packages ────────────────────────────── tidyverse 1.2.1 ─
#> ✔ ggplot2 3.1.0       ✔ purrr   0.2.5  
#> ✔ tibble  2.1.1       ✔ dplyr   0.8.0.1
#> ✔ tidyr   0.8.2       ✔ stringr 1.4.0  
#> ✔ readr   1.1.1       ✔ forcats 0.3.0
#> ─ Conflicts ─────────────────────────────── tidyverse_conflicts() ─
#> ✖ ggplot2::%+%()           masks psych::%+%()
#> ✖ ggplot2::alpha()         masks psych::alpha()
#> ✖ lubridate::as.difftime() masks base::as.difftime()
#> ✖ lubridate::date()        masks base::date()
#> ✖ dplyr::filter()          masks stats::filter()
#> ✖ lubridate::intersect()   masks base::intersect()
#> ✖ dplyr::lag()             masks stats::lag()
#> ✖ lubridate::setdiff()     masks base::setdiff()
#> ✖ lubridate::union()       masks base::union()
dataset %>% 
    describe()
#> Warning in FUN(newX[, i], ...): min里所有的参数都不存在; 回覆Inf
#> Warning in FUN(newX[, i], ...): max里所有的参数都不存在；回覆-Inf
#>            vars      n mean   sd median trimmed  mad   min   max range
#> short_name    1 183093 2.26 1.18   2.00    2.20 1.48  1.00  4.00  3.00
#> t             2 183093  NaN   NA     NA     NaN   NA   Inf  -Inf  -Inf
#> p             3 183093 4.85 9.50   0.00    2.26 0.17 -0.40 48.83 49.23
#> phat          4 183093 4.84 8.88   0.17    2.47 0.47 -3.55 46.97 50.52
#>            skew kurtosis   se
#> short_name 0.33    -1.40 0.00
#> t            NA       NA   NA
#> p          2.60     6.38 0.02
#> phat       2.39     5.11 0.02
phv_metric(
    id = dataset$short_name
    ,t = dataset$t
    ,y = dataset$p
    ,yhat = dataset$phat
)
#> [1] 0.1192559