Cubist models

Function Works
tidypredict_fit(), tidypredict_sql(), parse_model() ✔
tidypredict_to_column() ✔
tidypredict_test() ✗
tidypredict_interval(), tidypredict_sql_interval() ✗
parsnip ✗

tidypredict_ functions

library(Cubist)
data("BostonHousing", package = "mlbench")

model <- Cubist::cubist(
  x = BostonHousing[, -14],
  y = BostonHousing$medv,
  committees = 3
)

  • Create the R formula

    tidypredict_fit(model)
#> ((ifelse(nox > 0.66799998, -1.11 + crim * -0.02 + nox * 21.4 + 
#>     rm * 0.1 + age * -0.003 + dis * 2.93 + ptratio * -0.13 + 
#>     b * 0.008 + lstat * -0.33, 0) + ifelse(lstat > 9.5900002 & 
#>     nox <= 0.66799998, 23.57 + crim * 0.05 + nox * -5.2 + rm * 
#>     3.1 + age * -0.048 + dis * -0.81 + rad * 0.02 + tax * -0.0041 + 
#>     ptratio * -0.71 + b * 0.01 + lstat * -0.15, 0) + ifelse(lstat <= 
#>     9.5900002 & rm <= 6.2259998, 1.18 + crim * 3.83 + rm * 4.3 + 
#>     age * -0.06 + dis * -0.09 + tax * -0.003 + ptratio * -0.08 + 
#>     lstat * -0.11, 0) + ifelse(lstat <= 9.5900002 & rm > 6.2259998, 
#>     -4.71 + crim * 2.22 + zn * 0.008 + nox * -1.7 + rm * 9.2 + 
#>         age * -0.04 + dis * -0.71 + rad * 0.03 + tax * -0.0182 + 
#>         ptratio * -0.72 + lstat * -0.83, 0))/((nox > 0.66799998) + 
#>     (lstat > 9.5900002 & nox <= 0.66799998) + (lstat <= 9.5900002 & 
#>     rm <= 6.2259998) + (lstat <= 9.5900002 & rm > 6.2259998)) + 
#>     (ifelse(dis <= 1.7553999 & lstat > 5.1199999, 122.32 + crim * 
#>         -0.29 + nox * -21.6 + rm * -3 + dis * -30.88 + rad * 
#>         0.02 + tax * -0.001 + b * -0.023 + lstat * -0.73, 0) + 
#>         ifelse(rm <= 6.5450001 & lstat > 5.1199999, 27.8 + crim * 
#>             -0.16 + zn * 0.007 + nox * -3.9 + rm * 2 + age * 
#>             -0.035 + dis * -0.7 + rad * 0.28 + tax * -0.0135 + 
#>             ptratio * -0.6 + b * 0.013 + lstat * -0.25, 0) + 
#>         ifelse(rm > 6.5450001 & lstat > 5.1199999, 22.21 + crim * 
#>             -0.04 + zn * 0.01 + indus * -0.02 + nox * -4 + rm * 
#>             4.7 + dis * -0.34 + rad * 0.11 + tax * -0.0248 + 
#>             ptratio * -0.9 + b * 0.002 + lstat * -0.1, 0) + ifelse(lstat <= 
#>         5.1199999 & rm <= 8.0340004, -71.95 + rm * 17 + age * 
#>         -0.06 + tax * -0.0112 + ptratio * -0.48 + lstat * -0.03, 
#>         0) + ifelse(rm > 8.0340004 & dis > 3.1991999, -32.79 + 
#>         crim * -0.01 + zn * 0.005 + nox * -1.8 + rm * 12.9 + 
#>         age * -0.117 + dis * -0.15 + rad * 0.04 + tax * -0.0246 + 
#>         ptratio * -1.05 + lstat * -0.04, 0) + ifelse(lstat <= 
#>         5.1199999 & dis <= 3.1991999, 53.41 + rm * 1.6 + dis * 
#>         -7.16 + tax * 0.0088 + lstat * -0.68, 0))/((dis <= 1.7553999 & 
#>         lstat > 5.1199999) + (rm <= 6.5450001 & lstat > 5.1199999) + 
#>         (rm > 6.5450001 & lstat > 5.1199999) + (lstat <= 5.1199999 & 
#>         rm <= 8.0340004) + (rm > 8.0340004 & dis > 3.1991999) + 
#>         (lstat <= 5.1199999 & dis <= 3.1991999)) + (ifelse(nox > 
#>     0.66799998, -36.31 + crim * 0.08 + nox * 48.4 + dis * 7.52 + 
#>     b * 0.01 + lstat * -0.24, 0) + ifelse(lstat > 9.5299997 & 
#>     nox <= 0.66799998, 28.04 + nox * -4.8 + rm * 2.9 + age * 
#>     -0.051 + dis * -0.86 + rad * 0.01 + tax * -0.0019 + ptratio * 
#>     -0.72 + lstat * -0.12, 0) + ifelse(lstat <= 9.5299997, -26.05 + 
#>     crim * 0.89 + nox * -2.3 + rm * 9.6 + dis * -0.17 + rad * 
#>     0.02 + tax * -0.0055 + ptratio * -0.12 + b * 0.001 + lstat * 
#>     -0.74, 0) + ifelse(lstat <= 9.5299997 & dis <= 2.6403, 136.67 + 
#>     crim * 7.2 + nox * -96.6 + rm * 1.1 + tax * -0.0033 + ptratio * 
#>     -3.31 + lstat * -0.1, 0))/((nox > 0.66799998) + (lstat > 
#>     9.5299997 & nox <= 0.66799998) + (lstat <= 9.5299997) + (lstat <= 
#>     9.5299997 & dis <= 2.6403)))/3

  • SQL output example

    tidypredict_sql(model, dbplyr::simulate_odbc())
#> <SQL> ((((((CASE WHEN ("nox" > 0.66799998) THEN ((((((((-1.11 + "crim" * -0.02) + "nox" * 21.4) + "rm" * 0.1) + "age" * -0.003) + "dis" * 2.93) + "ptratio" * -0.13) + "b" * 0.008) + "lstat" * -0.33) WHEN NOT ("nox" > 0.66799998) THEN 0.0 END + CASE WHEN ("lstat" > 9.5900002 AND "nox" <= 0.66799998) THEN ((((((((((23.57 + "crim" * 0.05) + "nox" * -5.2) + "rm" * 3.1) + "age" * -0.048) + "dis" * -0.81) + "rad" * 0.02) + "tax" * -0.0041) + "ptratio" * -0.71) + "b" * 0.01) + "lstat" * -0.15) WHEN NOT ("lstat" > 9.5900002 AND "nox" <= 0.66799998) THEN 0.0 END) + CASE WHEN ("lstat" <= 9.5900002 AND "rm" <= 6.2259998) THEN (((((((1.18 + "crim" * 3.83) + "rm" * 4.3) + "age" * -0.06) + "dis" * -0.09) + "tax" * -0.003) + "ptratio" * -0.08) + "lstat" * -0.11) WHEN NOT ("lstat" <= 9.5900002 AND "rm" <= 6.2259998) THEN 0.0 END) + CASE WHEN ("lstat" <= 9.5900002 AND "rm" > 6.2259998) THEN ((((((((((-4.71 + "crim" * 2.22) + "zn" * 0.008) + "nox" * -1.7) + "rm" * 9.2) + "age" * -0.04) + "dis" * -0.71) + "rad" * 0.03) + "tax" * -0.0182) + "ptratio" * -0.72) + "lstat" * -0.83) WHEN NOT ("lstat" <= 9.5900002 AND "rm" > 6.2259998) THEN 0.0 END) / ((("nox" > 0.66799998 + "lstat" > 9.5900002 AND "nox" <= 0.66799998) + "lstat" <= 9.5900002 AND "rm" <= 6.2259998) + "lstat" <= 9.5900002 AND "rm" > 6.2259998)) + (((((CASE WHEN ("dis" <= 1.7553999 AND "lstat" > 5.1199999) THEN ((((((((122.32 + "crim" * -0.29) + "nox" * -21.6) + "rm" * -3.0) + "dis" * -30.88) + "rad" * 0.02) + "tax" * -0.001) + "b" * -0.023) + "lstat" * -0.73) WHEN NOT ("dis" <= 1.7553999 AND "lstat" > 5.1199999) THEN 0.0 END + CASE WHEN ("rm" <= 6.5450001 AND "lstat" > 5.1199999) THEN (((((((((((27.8 + "crim" * -0.16) + "zn" * 0.007) + "nox" * -3.9) + "rm" * 2.0) + "age" * -0.035) + "dis" * -0.7) + "rad" * 0.28) + "tax" * -0.0135) + "ptratio" * -0.6) + "b" * 0.013) + "lstat" * -0.25) WHEN NOT ("rm" <= 6.5450001 AND "lstat" > 5.1199999) THEN 0.0 END) + CASE WHEN ("rm" > 6.5450001 AND "lstat" > 5.1199999) THEN (((((((((((22.21 + "crim" * -0.04) + "zn" * 0.01) + "indus" * -0.02) + "nox" * -4.0) + "rm" * 4.7) + "dis" * -0.34) + "rad" * 0.11) + "tax" * -0.0248) + "ptratio" * -0.9) + "b" * 0.002) + "lstat" * -0.1) WHEN NOT ("rm" > 6.5450001 AND "lstat" > 5.1199999) THEN 0.0 END) + CASE WHEN ("lstat" <= 5.1199999 AND "rm" <= 8.0340004) THEN (((((-71.95 + "rm" * 17.0) + "age" * -0.06) + "tax" * -0.0112) + "ptratio" * -0.48) + "lstat" * -0.03) WHEN NOT ("lstat" <= 5.1199999 AND "rm" <= 8.0340004) THEN 0.0 END) + CASE WHEN ("rm" > 8.0340004 AND "dis" > 3.1991999) THEN ((((((((((-32.79 + "crim" * -0.01) + "zn" * 0.005) + "nox" * -1.8) + "rm" * 12.9) + "age" * -0.117) + "dis" * -0.15) + "rad" * 0.04) + "tax" * -0.0246) + "ptratio" * -1.05) + "lstat" * -0.04) WHEN NOT ("rm" > 8.0340004 AND "dis" > 3.1991999) THEN 0.0 END) + CASE WHEN ("lstat" <= 5.1199999 AND "dis" <= 3.1991999) THEN ((((53.41 + "rm" * 1.6) + "dis" * -7.16) + "tax" * 0.0088) + "lstat" * -0.68) WHEN NOT ("lstat" <= 5.1199999 AND "dis" <= 3.1991999) THEN 0.0 END) / ((((("dis" <= 1.7553999 AND "lstat" > 5.1199999 + "rm" <= 6.5450001 AND "lstat" > 5.1199999) + "rm" > 6.5450001 AND "lstat" > 5.1199999) + "lstat" <= 5.1199999 AND "rm" <= 8.0340004) + "rm" > 8.0340004 AND "dis" > 3.1991999) + "lstat" <= 5.1199999 AND "dis" <= 3.1991999)) + (((CASE WHEN ("nox" > 0.66799998) THEN (((((-36.31 + "crim" * 0.08) + "nox" * 48.4) + "dis" * 7.52) + "b" * 0.01) + "lstat" * -0.24) WHEN NOT ("nox" > 0.66799998) THEN 0.0 END + CASE WHEN ("lstat" > 9.5299997 AND "nox" <= 0.66799998) THEN ((((((((28.04 + "nox" * -4.8) + "rm" * 2.9) + "age" * -0.051) + "dis" * -0.86) + "rad" * 0.01) + "tax" * -0.0019) + "ptratio" * -0.72) + "lstat" * -0.12) WHEN NOT ("lstat" > 9.5299997 AND "nox" <= 0.66799998) THEN 0.0 END) + CASE WHEN ("lstat" <= 9.5299997) THEN (((((((((-26.05 + "crim" * 0.89) + "nox" * -2.3) + "rm" * 9.6) + "dis" * -0.17) + "rad" * 0.02) + "tax" * -0.0055) + "ptratio" * -0.12) + "b" * 0.001) + "lstat" * -0.74) WHEN NOT ("lstat" <= 9.5299997) THEN 0.0 END) + CASE WHEN ("lstat" <= 9.5299997 AND "dis" <= 2.6403) THEN ((((((136.67 + "crim" * 7.2) + "nox" * -96.6) + "rm" * 1.1) + "tax" * -0.0033) + "ptratio" * -3.31) + "lstat" * -0.1) WHEN NOT ("lstat" <= 9.5299997 AND "dis" <= 2.6403) THEN 0.0 END) / ((("nox" > 0.66799998 + "lstat" > 9.5299997 AND "nox" <= 0.66799998) + "lstat" <= 9.5299997) + "lstat" <= 9.5299997 AND "dis" <= 2.6403)) / 3.0

  • Add the prediction to the original table

    library(dplyr)

BostonHousing %>%
  tidypredict_to_column(model) %>%
  glimpse()
#> Rows: 506
#> Columns: 15
#> $ crim    <dbl> 0.00632, 0.02731, 0.02729, 0.03237, 0.06905, 0.02985, 0.08829,…
#> $ zn      <dbl> 18.0, 0.0, 0.0, 0.0, 0.0, 0.0, 12.5, 12.5, 12.5, 12.5, 12.5, 1…
#> $ indus   <dbl> 2.31, 7.07, 7.07, 2.18, 2.18, 2.18, 7.87, 7.87, 7.87, 7.87, 7.…
#> $ chas    <fct> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
#> $ nox     <dbl> 0.538, 0.469, 0.469, 0.458, 0.458, 0.458, 0.524, 0.524, 0.524,…
#> $ rm      <dbl> 6.575, 6.421, 7.185, 6.998, 7.147, 6.430, 6.012, 6.172, 5.631,…
#> $ age     <dbl> 65.2, 78.9, 61.1, 45.8, 54.2, 58.7, 66.6, 96.1, 100.0, 85.9, 9…
#> $ dis     <dbl> 4.0900, 4.9671, 4.9671, 6.0622, 6.0622, 6.0622, 5.5605, 5.9505…
#> $ rad     <dbl> 1, 2, 2, 3, 3, 3, 5, 5, 5, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4,…
#> $ tax     <dbl> 296, 242, 242, 222, 222, 222, 311, 311, 311, 311, 311, 311, 31…
#> $ ptratio <dbl> 15.3, 17.8, 17.8, 18.7, 18.7, 18.7, 15.2, 15.2, 15.2, 15.2, 15…
#> $ b       <dbl> 396.90, 396.90, 392.83, 394.63, 396.90, 394.12, 395.60, 396.90…
#> $ lstat   <dbl> 4.98, 9.14, 4.03, 2.94, 5.33, 5.21, 12.43, 19.15, 29.93, 17.10…
#> $ medv    <dbl> 24.0, 21.6, 34.7, 33.4, 36.2, 28.7, 22.9, 27.1, 16.5, 18.9, 15…
#> $ fit     <dbl> 27.50665, 22.71805, 34.78128, 33.19372, 31.93653, 25.03739, 21…

We are not able to give an exact match of the original predictions due to a minor bug in Cubist.

Parse model spec

Here is an example of the model spec:

pm <- parse_model(model)
str(pm, 2)
#> List of 2
#>  $ general:List of 6
#>   ..$ model       : chr "cubist"
#>   ..$ type        : chr "tree"
#>   ..$ version     : num 3
#>   ..$ mode        : chr "ifelse"
#>   ..$ n_committees: num 3
#>   ..$ ommittee_id : int [1:14] 1 1 1 1 2 2 2 2 2 2 ...
#>  $ trees  :List of 1
#>   ..$ :List of 14
#>  - attr(*, "class")= chr [1:3] "parsed_model" "pm_tree" "list"
str(pm$terms[1:2])
#>  NULL

Limitations

  • tidypredict_test() is not supported
  • Prediction intervals are not supported
  • Cubist uses 32-bit floats internally, which may cause prediction discrepancies at exact split boundaries. See the float precision article for details.