AQI {CASMI}R Documentation

AQI Index

Description

A quantitative measure of dataset quality. The AQI Index score indicates the degree that how features are associated with the outcome in a dataset. (synonyms of "feature": "variable" "factor" "attribute")
For more information, please refer to the corresponding publication: Shi, J., Zhang, J. and Ge, Y. (2019), "An Association-Based Intrinsic Quality Index for Healthcare Dataset Ranking" <doi:10.1109/ICHI.2019.8904553>

Usage

AQI(data, alpha.filter = 0.2)

Arguments

data

data frame (features as columns and observations as rows). The outcome variable (Y) MUST be the last column. It requires at least one features and only one outcome. Both the features (Xs) and the outcome (Y) MUST be discrete (if not naturally discrete, you may try the 'autoBin.binary' function in the same package).

alpha.filter

level of significance for the mutual information test of independence in step 2 (<doi:10.1109/ICHI.2019.8904553>). By default, 'alpha.filter = 0.2'.

Value

The AQI Index score.

Examples

## Generate a toy dataset: "data"
n <- 10000
set.seed(1)
x1 <- rbinom(n, 3, 0.5) + 0.2
set.seed(2)
x2 <- rbinom(n, 2, 0.8) + 0.5
set.seed(3)
x3 <- rbinom(n, 5, 0.3)
set.seed(4)
error <- round(runif(n, min=-1, max=1))
y <- x1 + x3 + error
data <- data.frame(cbind(x1, x2, x3, y))
colnames(data) <- c("feature1", "feature2", "feature3", "Y")

## Calculate the AQI score of "data"
AQI(data)

[Package CASMI version 1.2.2 Index]