plt.single.class.importance: Single class importance plot

Description Usage Arguments

Description

Plot the feature importance per state. Feature importance is normalized.
Specify colmin to obtain a plot containing all coordinates with a feature importance larger than this minimum importance.
Specify nfeatures to obtain a plot containing the n most important coordinates. TODO update parameter descriptions

Usage

1
2
3
plt.single.class.importance(pre = "singleClassImportance/sci", model = NA,
  names = NA, states = NA, colmin = NA, nfeatures = NA, pdim = 10,
  width = 1)

Arguments

pre

prefix for saved plots

model

directory to xgb.model file. If NA (default) model from ./model/xgb.model is taken.

names

custom feature (coordinate) names file

states

custom states file (default=NA : states are numbered consecutively starting from 1)

colmin

Sets minimum of imnportance a feature has to contribute to at least one class. Can be set as vector to produce multiple plots: colmin = c(0.1,0.2,0.3)

nfeatures

Returns plot with 'nfeatures' most important features. Can be set as vector to produce multiple plots: c(4,5,6)

pdim

plot dimension (height = pdim)

width

plot width (pdim*width)


sbbrandt/xgbAnalysis documentation built on May 29, 2019, 9:11 a.m.