performance_bin: Diagnose Performance Binned Variable

In choonghyunryu/dlookr: Tools for Data Diagnosis, Exploration, Transformation

Diagnose Performance Binned Variable

Description

The performance_bin() calculates metrics to evaluate the performance of binned variable for binomial classification model.

Usage

performance_bin(y, x, na.rm = FALSE)

Arguments

y	character or numeric, integer, factor. a binary response variable (0, 1). The variable must contain only the integers 0 and 1 as element. However, in the case of factor/character having two levels, it is performed while type conversion is performed in the calculation process.
x	integer or factor, character. At least 2 different values. and Inf is not allowed.
na.rm	logical. a logical indicating whether missing values should be removed.

Details

This function is useful when used with the mutate/transmute function of the dplyr package.

Value

an object of "performance_bin" class. vaue of data.frame is as follows.

• Bin : character. bins.

• CntRec : integer. frequency by bins.

• CntPos : integer. frequency of positive by bins.

• CntNeg : integer. frequency of negative by bins.

• CntCumPos : integer. cumulate frequency of positive by bins.

• CntCumNeg : integer. cumulate frequency of negative by bins.

• RatePos : integer. relative frequency of positive by bins.

• RateNeg : integer. relative frequency of negative by bins.

• RateCumPos : numeric. cumulate relative frequency of positive by bins.

• RateCumNeg : numeric. cumulate relative frequency of negative by bins.

• Odds : numeric. odd ratio.

• LnOdds : numeric. loged odd ratio.

• WoE : numeric. weight of evidence.

• IV : numeric. Jeffrey's Information Value.

• JSD : numeric. Jensen-Shannon Divergence.

• AUC : numeric. AUC. area under curve.

Attributes of "performance_bin" class is as follows.

• names : character. variable name of data.frame with "Binning Table".

• class : character. name of class. "performance_bin" "data.frame".

• row.names : character. row name of data.frame with "Binning Table".

• IV : numeric. Jeffrey's Information Value.

• JSD : numeric. Jensen-Shannon Divergence.

• KS : numeric. Kolmogorov-Smirnov Statistics.

• gini : numeric. Gini index.

• HHI : numeric. Herfindahl-Hirschman Index.

• HHI_norm : numeric.normalized Herfindahl-Hirschman Index.

• Cramer_V : numeric. Cramer's V Statistics.

• chisq_test : data.frame. table of significance tests. name is as follows.

  • Bin A : character. first bins.

  • Bin B : character. second bins.

  • statistics : numeric. statistics of Chi-square test.

  • p_value : numeric. p-value of Chi-square test.

See Also

summary.performance_bin, plot.performance_bin, binning_by.

Examples

# Generate data for the example
heartfailure2 <- heartfailure

set.seed(123)
heartfailure2[sample(seq(NROW(heartfailure2)), 5), "creatinine"] <- NA

# Change the target variable to 0(negative) and 1(positive).
heartfailure2$death_event_2 <- ifelse(heartfailure2$death_event %in% "Yes", 1, 0)

# Binnig from creatinine to platelets_bin.
breaks <- c(0,  1,  2, 10)
heartfailure2$creatinine_bin <- cut(heartfailure2$creatinine, breaks)

# Diagnose performance binned variable
perf <- performance_bin(heartfailure2$death_event_2, heartfailure2$creatinine_bin) 
perf
summary(perf)

plot(perf)

# Diagnose performance binned variable without NA
perf <- performance_bin(heartfailure2$death_event_2, heartfailure2$creatinine_bin, na.rm = TRUE) 
perf
summary(perf)

plot(perf)

choonghyunryu/dlookr documentation built on June 11, 2024, 9:12 a.m.