Skip to main content

Table 3 Random forest classification models of air/surface

From: The subway microbiome: seasonal dynamics and direct comparison of air and surface bacterial communities

Air/surface

Out-of-bag estimate of error rate: 6.1%

 Confusion matrix

Air

Surface

Class error (%)

  Air

56

13

18.8

  Surface

2

175

1.1

Most important genera in sample classification

 Family:Genera

Air

Surface

MDA

  Burkholderiaceae:Ralstonia

0.027

0.010

0.015

  Streptomycetaceae:Streptomyces

0.020

0.006

0.010

  Pseudonocardiaceae:Pseudonocardia

0.018

0.006

0.009

  Streptococcaceae:Streptococcus

0.015

0.004

0.007

  Pseudonocardiaceae:Saccharopolyspora

0.012

0.004

0.006

  Neisseriaceae:Neisseria

0.013

0.003

0.006

  Nocardiopsaceae:Nocardiopsis

0.011

0.004

0.006

  Rubrobacteriaceae:Rubrobacter

0.011

0.004

0.006

  Micrococcaceae:Micrococcus

0.008

0.004

0.005

  Carnobacteriaceae:Granulicatella

0.012

0.003

0.005

  Pasteurellaceae:Haemophilus

0.012

0.002

0.005

  Peptostreptococcaceae:Terrisporobacter

0.008

0.004

0.005

  Micrococcaceae:Pseudarthrobacter

0.009

0.003

0.005

  Planococcaceae:Planomicrobium

0.009

0.003

0.005

  Halococcaceae:Halococcus

0.007

0.003

0.004

  Micrococcaceae:Rothia

0.008

0.002

0.004

  Halococcaceae:Halalkalicoccus

0.007

0.002

0.003

  Porphyromonadaceae:Porphyromonas

0.007

0.002

0.003

  Planococcaceae:Planococcus

0.007

0.002

0.003

  Pseudonocardiaceae:Actinomycetospora

0.006

0.002

0.003

  1. Confusion matrices show the classification of samples and the associated class error. The mean decrease in model accuracy (MDA; from removing the genus in question) and mean Z-scores are given for the 20 most important genera for classifying samples