Obtain rules as a tidy tibble from a C5.0 model

Source: R/c5.R

Each row corresponds to a rule. A rule can be copied into `dplyr::filter` to filter the observations corresponding to a rule

# S3 method for C5.0
tidyRules(object, ...)

Arguments

object

Fitted model object with rules
...

Other arguments (See details)

Value

A tibble where each row corresponds to a rule. The columns are: support, confidence, lift, lhs, rhs, n_conditions

Details

Optional named arguments:

  • laplace(flag, default: TRUE) is supported. This computes confidence with laplace correction as documented under 'Rulesets' here: [C5 doc](https://www.rulequest.com/see5-unix.html).

  • language (string, default: "r"): language where the rules are parsable. The allowed options is one among: r, python, sql

Examples

data("attrition", package = "modeldata")
attrition <- tibble::as_tibble(attrition)
c5_model <- C50::C5.0(Attrition ~., data = attrition, rules = TRUE)
summary(c5_model)
#> 
#> Call:
#> C5.0.formula(formula = Attrition ~ ., data = attrition, rules = TRUE)
#> 
#> 
#> C5.0 [Release 2.07 GPL Edition]  	Thu Jun  4 17:35:29 2020
#> -------------------------------
#> 
#> Class specified by attribute `outcome'
#> 
#> Read 1470 cases (31 attributes) from undefined.data
#> 
#> Rules:
#> 
#> Rule 1: (521/30, lift 1.1)
#> 	EnvironmentSatisfaction in [Medium-Very_High]
#> 	JobInvolvement in [Medium-Very_High]
#> 	OverTime = No
#> 	TrainingTimesLastYear > 1
#> 	WorkLifeBalance in [Better-Best]
#> 	->  class No  [0.941]
#> 
#> Rule 2: (195/14, lift 1.1)
#> 	JobRole = Research_Scientist
#> 	OverTime = No
#> 	->  class No  [0.924]
#> 
#> Rule 3: (1347/183, lift 1.0)
#> 	TotalWorkingYears > 2
#> 	->  class No  [0.864]
#> 
#> Rule 4: (16, lift 5.9)
#> 	JobLevel <= 1
#> 	MonthlyIncome <= 2468
#> 	OverTime = Yes
#> 	TotalWorkingYears > 2
#> 	YearsAtCompany <= 3
#> 	->  class Yes  [0.944]
#> 
#> Rule 5: (13, lift 5.8)
#> 	DailyRate <= 722
#> 	JobLevel <= 1
#> 	MonthlyIncome <= 2468
#> 	OverTime = Yes
#> 	TotalWorkingYears > 2
#> 	->  class Yes  [0.933]
#> 
#> Rule 6: (9, lift 5.6)
#> 	EnvironmentSatisfaction in [Low-Medium]
#> 	MaritalStatus in {Divorced, Married}
#> 	NumCompaniesWorked > 4
#> 	OverTime = Yes
#> 	PerformanceRating = Excellent
#> 	RelationshipSatisfaction in [Low-High]
#> 	->  class Yes  [0.909]
#> 
#> Rule 7: (9, lift 5.6)
#> 	EnvironmentSatisfaction in [Low-Medium]
#> 	Gender = Male
#> 	MaritalStatus in {Divorced, Married}
#> 	NumCompaniesWorked > 4
#> 	OverTime = Yes
#> 	PerformanceRating = Excellent
#> 	->  class Yes  [0.909]
#> 
#> Rule 8: (6, lift 5.4)
#> 	JobRole = Laboratory_Technician
#> 	MaritalStatus = Single
#> 	MonthlyIncome > 2468
#> 	OverTime = Yes
#> 	TrainingTimesLastYear <= 2
#> 	->  class Yes  [0.875]
#> 
#> Rule 9: (14/1, lift 5.4)
#> 	JobRole in {Laboratory_Technician, Sales_Representative}
#> 	MonthlyIncome <= 2657
#> 	TotalWorkingYears <= 2
#> 	WorkLifeBalance in [Bad-Good]
#> 	->  class Yes  [0.875]
#> 
#> Rule 10: (13/1, lift 5.4)
#> 	Department = Sales
#> 	MaritalStatus = Single
#> 	OverTime = Yes
#> 	YearsSinceLastPromotion > 1
#> 	->  class Yes  [0.867]
#> 
#> Rule 11: (5, lift 5.3)
#> 	Age <= 44
#> 	OverTime = No
#> 	TotalWorkingYears > 2
#> 	WorkLifeBalance = Bad
#> 	YearsWithCurrManager <= 0
#> 	->  class Yes  [0.857]
#> 
#> Rule 12: (5, lift 5.3)
#> 	EducationField = Life_Sciences
#> 	EnvironmentSatisfaction in [Low-Medium]
#> 	JobInvolvement = Low
#> 	MonthlyIncome > 2468
#> 	OverTime = Yes
#> 	->  class Yes  [0.857]
#> 
#> Rule 13: (5, lift 5.3)
#> 	JobInvolvement = Low
#> 	OverTime = No
#> 	TotalWorkingYears <= 2
#> 	->  class Yes  [0.857]
#> 
#> Rule 14: (5, lift 5.3)
#> 	Department = Sales
#> 	JobInvolvement in [Low-Medium]
#> 	JobSatisfaction in [Medium-High]
#> 	MaritalStatus = Single
#> 	OverTime = Yes
#> 	YearsSinceLastPromotion <= 1
#> 	->  class Yes  [0.857]
#> 
#> Rule 15: (4, lift 5.2)
#> 	JobRole in {Laboratory_Technician, Sales_Representative}
#> 	TotalWorkingYears <= 2
#> 	TrainingTimesLastYear <= 1
#> 	->  class Yes  [0.833]
#> 
#> Rule 16: (4, lift 5.2)
#> 	JobSatisfaction = High
#> 	OverTime = No
#> 	WorkLifeBalance = Bad
#> 	YearsSinceLastPromotion > 6
#> 	->  class Yes  [0.833]
#> 
#> Rule 17: (4, lift 5.2)
#> 	Department = Sales
#> 	JobInvolvement = Very_High
#> 	MaritalStatus = Single
#> 	OverTime = Yes
#> 	->  class Yes  [0.833]
#> 
#> Rule 18: (8/1, lift 5.0)
#> 	EnvironmentSatisfaction = Low
#> 	JobRole = Laboratory_Technician
#> 	TotalWorkingYears <= 2
#> 	->  class Yes  [0.800]
#> 
#> Rule 19: (6/1, lift 4.7)
#> 	JobRole = Human_Resources
#> 	TotalWorkingYears <= 2
#> 	->  class Yes  [0.750]
#> 
#> Rule 20: (35/10, lift 4.4)
#> 	OverTime = Yes
#> 	TotalWorkingYears <= 2
#> 	->  class Yes  [0.703]
#> 
#> Rule 21: (7/2, lift 4.1)
#> 	OverTime = No
#> 	TotalWorkingYears > 2
#> 	WorkLifeBalance = Bad
#> 	YearsWithCurrManager <= 0
#> 	->  class Yes  [0.667]
#> 
#> Rule 22: (49/17, lift 4.0)
#> 	Department = Sales
#> 	MaritalStatus = Single
#> 	OverTime = Yes
#> 	->  class Yes  [0.647]
#> 
#> Rule 23: (6/2, lift 3.9)
#> 	OverTime = No
#> 	WorkLifeBalance = Bad
#> 	YearsSinceLastPromotion > 6
#> 	->  class Yes  [0.625]
#> 
#> Rule 24: (57/31, lift 2.8)
#> 	EnvironmentSatisfaction in [Low-Medium]
#> 	MonthlyRate > 16620
#> 	OverTime = Yes
#> 	->  class Yes  [0.458]
#> 
#> Default class: No
#> 
#> 
#> Evaluation on training data (1470 cases):
#> 
#> 	        Rules     
#> 	  ----------------
#> 	    No      Errors
#> 
#> 	    24  134( 9.1%)   <<
#> 
#> 
#> 	   (a)   (b)    <-classified as
#> 	  ----  ----
#> 	  1218    15    (a): class No
#> 	   119   118    (b): class Yes
#> 
#> 
#> 	Attribute usage:
#> 
#> 	 95.71%	TotalWorkingYears
#> 	 54.22%	OverTime
#> 	 40.20%	EnvironmentSatisfaction
#> 	 37.28%	WorkLifeBalance
#> 	 36.73%	JobInvolvement
#> 	 36.12%	TrainingTimesLastYear
#> 	 15.51%	JobRole
#> 	  4.56%	MaritalStatus
#> 	  3.88%	MonthlyRate
#> 	  3.33%	Department
#> 	  3.33%	MonthlyIncome
#> 	  1.63%	JobLevel
#> 	  1.63%	YearsSinceLastPromotion
#> 	  1.09%	YearsAtCompany
#> 	  0.88%	DailyRate
#> 	  0.82%	NumCompaniesWorked
#> 	  0.82%	PerformanceRating
#> 	  0.61%	Gender
#> 	  0.61%	JobSatisfaction
#> 	  0.61%	RelationshipSatisfaction
#> 	  0.48%	YearsWithCurrManager
#> 	  0.34%	Age
#> 	  0.34%	EducationField
#> 
#> 
#> Time: 0.0 secs
#> 
tidyRules(c5_model)
#> # A tibble: 24 x 8
#>       id LHS             RHS   support confidence  lift rule_number trial_number
#>    <int> <chr>           <chr>   <int>      <dbl> <dbl>       <int>        <int>
#>  1     1 EnvironmentSat… No        521      0.941   1.1           1            1
#>  2     2 JobRole == 'Re… No        195      0.924   1.1           2            1
#>  3     3 TotalWorkingYe… No       1347      0.864   1             3            1
#>  4     4 JobLevel <= 1 … Yes        16      0.944   5.9           4            1
#>  5     5 DailyRate <= 7… Yes        13      0.933   5.8           5            1
#>  6     6 EnvironmentSat… Yes         9      0.909   5.6           6            1
#>  7     7 EnvironmentSat… Yes         9      0.909   5.6           7            1
#>  8     8 JobRole == 'La… Yes         6      0.875   5.4           8            1
#>  9     9 JobRole %in% c… Yes        14      0.875   5.4           9            1
#> 10    10 Department == … Yes        13      0.867   5.4          10            1
#> # … with 14 more rows