Extract predicted components from a Random Planted Forest
Source:R/predict_components.R
predict_components.RdPrediction components are a functional decomposition of the model prediction. The sum of all components equals the overall predicted value for an observation.
Arguments
- object
[rpf]: A fittedrpfmodel.- new_data
[data.frame | matrix]: New observations to predict.- max_interaction
[integer(1) | NULL: NULL]: Maximum degree of interactions to consider.NULLuses themax_interactionparameter from therpfobject. Must be between1(main effects only) and themax_interactionof therpfobject.- predictors
[character | NULL: NULL]: Vector of one or more column names of predictor variables innew_datato extract components for. IfNULL, all variables and their interactions are returned.
Value
A list with elements:
m(data.table): Components for each main effect and interaction term, representing the functional decomposition of the prediction. All components together with the intercept sum up to the prediction. For multiclass classification, the number of output columns is multiplied by the number of levels in the outcome.intercept(numeric): Expected value of the prediction. For multiclass classification, one value per target level, named bytarget_levels.x(data.table): Copy ofnew_datacontaining predictors selected bypredictors.target_levels(character): For multiclass classification only: Vector of target levels which can be used to disassemblem, as names include both term and target level.
Details
Extracts all possible components up to max_interaction degrees,
up to the value set when calling rpf(). The intercept is always included.
Optionally predictors can be specified to only include components including the given variables.
If max_interaction is greater than length(predictors), the max_interaction will be lowered accordingly.
The components are only unique for a purified forest.
An unpurified forest is purified in place on first use with the defaults of purify().
For other settings, such as more threads, call purify() before predict_components().
Note
Depending on the number of predictors and max_interaction, the number of components will
increase drastically to sum(choose(ncol(new_data), seq_len(max_interaction))).
Examples
# Regression task, only some predictors
train <- mtcars[1:20, 1:4]
test <- mtcars[21:32, 1:4]
set.seed(23)
rpfit <- rpf(mpg ~ ., data = train, max_interaction = 3, ntrees = 30)
# Extract all components, including main effects and interaction terms up to `max_interaction`
(components <- predict_components(rpfit, test))
#> $m
#> cyl disp hp cyl:disp cyl:hp disp:hp
#> <num> <num> <num> <num> <num> <num>
#> 1: 0.9013797 3.2511514 1.3776657 0.08511109 -0.002332250 -0.11682939
#> 2: -0.6706529 -0.9186116 -0.4097990 -0.14816938 -0.079604722 0.10864962
#> 3: -0.6706529 -0.9186116 -0.4097990 -0.14816938 -0.079604722 0.10864962
#> 4: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 5: -0.6706529 -1.1532004 -0.3559023 -0.07494483 -0.084604722 0.12038493
#> 6: 0.9013797 4.6372466 4.7893287 0.17994554 -0.091148415 0.36694210
#> 7: 0.9013797 3.2511514 4.7893287 0.08511109 -0.091148415 0.07237951
#> 8: 0.9013797 4.6372466 0.6132023 0.17994554 0.006048960 -0.46138786
#> 9: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 10: -0.1689229 0.9540856 -0.3559023 -0.08618539 -0.045141470 -0.09680110
#> 11: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 12: 0.9013797 3.2511514 0.9815403 0.08511109 -0.005884567 0.03742733
#> cyl:disp:hp
#> <num>
#> 1: -0.06166953
#> 2: 0.04480135
#> 3: 0.04480135
#> 4: 0.07122433
#> 5: 0.06090002
#> 6: -0.13176489
#> 7: -0.10430258
#> 8: -0.02536408
#> 9: 0.07122433
#> 10: -0.01180950
#> 11: 0.07122433
#> 12: -0.06367401
#>
#> $intercept
#> [1] 20.49833
#>
#> $x
#> cyl disp hp
#> <num> <num> <num>
#> 1: 4 120.1 97
#> 2: 8 318.0 150
#> 3: 8 304.0 150
#> 4: 8 350.0 245
#> 5: 8 400.0 175
#> 6: 4 79.0 66
#> 7: 4 120.3 91
#> 8: 4 95.1 113
#> 9: 8 351.0 264
#> 10: 6 145.0 175
#> 11: 8 301.0 335
#> 12: 4 121.0 109
#>
#> attr(,"class")
#> [1] "glex" "rpf_components" "list"
# sums to prediction
cbind(
m_sum = rowSums(components$m) + components$intercept,
prediction = predict(rpfit, test)
)
#> m_sum .pred
#> 1 25.93281 25.93281
#> 2 18.42495 18.42495
#> 3 18.42495 18.42495
#> 4 15.06680 15.06680
#> 5 18.34031 18.34031
#> 6 31.15026 31.15026
#> 7 29.40223 29.40223
#> 8 26.34941 26.34941
#> 9 15.06680 15.06680
#> 10 20.68766 20.68766
#> 11 15.06680 15.06680
#> 12 25.68539 25.68539
# Only get components with interactions of a lower degree, ignoring 3-way interactions
predict_components(rpfit, test, max_interaction = 2)
#> $m
#> cyl disp hp cyl:disp cyl:hp disp:hp
#> <num> <num> <num> <num> <num> <num>
#> 1: 0.9013797 3.2511514 1.3776657 0.08511109 -0.002332250 -0.11682939
#> 2: -0.6706529 -0.9186116 -0.4097990 -0.14816938 -0.079604722 0.10864962
#> 3: -0.6706529 -0.9186116 -0.4097990 -0.14816938 -0.079604722 0.10864962
#> 4: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 5: -0.6706529 -1.1532004 -0.3559023 -0.07494483 -0.084604722 0.12038493
#> 6: 0.9013797 4.6372466 4.7893287 0.17994554 -0.091148415 0.36694210
#> 7: 0.9013797 3.2511514 4.7893287 0.08511109 -0.091148415 0.07237951
#> 8: 0.9013797 4.6372466 0.6132023 0.17994554 0.006048960 -0.46138786
#> 9: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 10: -0.1689229 0.9540856 -0.3559023 -0.08618539 -0.045141470 -0.09680110
#> 11: -0.6706529 -0.9186116 -3.6554527 -0.14816938 -0.017505790 -0.09236340
#> 12: 0.9013797 3.2511514 0.9815403 0.08511109 -0.005884567 0.03742733
#>
#> $intercept
#> [1] 20.49833
#>
#> $x
#> cyl disp hp
#> <num> <num> <num>
#> 1: 4 120.1 97
#> 2: 8 318.0 150
#> 3: 8 304.0 150
#> 4: 8 350.0 245
#> 5: 8 400.0 175
#> 6: 4 79.0 66
#> 7: 4 120.3 91
#> 8: 4 95.1 113
#> 9: 8 351.0 264
#> 10: 6 145.0 175
#> 11: 8 301.0 335
#> 12: 4 121.0 109
#>
#> $remainder
#> [1] -0.06166953 0.04480135 0.04480135 0.07122433 0.06090002 -0.13176489
#> [7] -0.10430258 -0.02536408 0.07122433 -0.01180950 0.07122433 -0.06367401
#>
#> attr(,"class")
#> [1] "glex" "rpf_components" "list"
# Only retrieve main effects
(main_effects <- predict_components(rpfit, test, max_interaction = 1))
#> $m
#> cyl disp hp
#> <num> <num> <num>
#> 1: 0.9013797 3.2511514 1.3776657
#> 2: -0.6706529 -0.9186116 -0.4097990
#> 3: -0.6706529 -0.9186116 -0.4097990
#> 4: -0.6706529 -0.9186116 -3.6554527
#> 5: -0.6706529 -1.1532004 -0.3559023
#> 6: 0.9013797 4.6372466 4.7893287
#> 7: 0.9013797 3.2511514 4.7893287
#> 8: 0.9013797 4.6372466 0.6132023
#> 9: -0.6706529 -0.9186116 -3.6554527
#> 10: -0.1689229 0.9540856 -0.3559023
#> 11: -0.6706529 -0.9186116 -3.6554527
#> 12: 0.9013797 3.2511514 0.9815403
#>
#> $intercept
#> [1] 20.49833
#>
#> $x
#> cyl disp hp
#> <num> <num> <num>
#> 1: 4 120.1 97
#> 2: 8 318.0 150
#> 3: 8 304.0 150
#> 4: 8 350.0 245
#> 5: 8 400.0 175
#> 6: 4 79.0 66
#> 7: 4 120.3 91
#> 8: 4 95.1 113
#> 9: 8 351.0 264
#> 10: 6 145.0 175
#> 11: 8 301.0 335
#> 12: 4 121.0 109
#>
#> $remainder
#> [1] -0.09572008 -0.07432313 -0.07432313 -0.18681424 0.02173540 0.32397433
#> [7] -0.03796040 -0.30075744 -0.18681424 -0.23993745 -0.18681424 0.05297984
#>
#> attr(,"class")
#> [1] "glex" "rpf_components" "list"
# The difference is the combined contribution of interaction effects
cbind(
m_sum = rowSums(main_effects$m) + main_effects$intercept,
prediction = predict(rpfit, test)
)
#> m_sum .pred
#> 1 26.02853 25.93281
#> 2 18.49927 18.42495
#> 3 18.49927 18.42495
#> 4 15.25362 15.06680
#> 5 18.31858 18.34031
#> 6 30.82629 31.15026
#> 7 29.44019 29.40223
#> 8 26.65016 26.34941
#> 9 15.25362 15.06680
#> 10 20.92759 20.68766
#> 11 15.25362 15.06680
#> 12 25.63241 25.68539