plot
Generates plots from flow records and fitted models (requires pandas and scipy).
usage: python3 -m flow_models.plot [-h] [--format {png,pdf}] [--single]
[--no-normalize] [--fft]
[-P {points,hist,kde,comp,comp_stack,comp_labels}]
[-C {comp,comp_stack,comp_labels}]
[-x {length,size,duration,rate}]
histogram [mixture]
Positional Arguments
- histogram
csv_hist file to plot
- mixture
mixture directory to plot
Named Arguments
- --format
Possible choices: png, pdf
plot file format
Default:
'png'- --single
plot PDF and CDF in single file
Default:
False- --no-normalize
do not normalize PDF datapoints
Default:
True- --fft
use FFT for calculating KDE
Default:
False- -P
Possible choices: points, hist, kde, comp, comp_stack, comp_labels
additional PDF plot modes (can be specified multiple times)
Default:
[]- -C
Possible choices: comp, comp_stack, comp_labels
additional CDF plot modes (can be specified multiple times)
Default:
[]- -x
Possible choices: length, size, duration, rate
x axis value
Default:
'length'
An important part of any modeling task is the visualization of both input data and resulting models. The plot tool can be used for that purpose. It can generate probability density (PDF), cumulative distribution function (CDF) and average packet size and packet interarrival time plots. It takes CSV histogram files and mixture model JSON files as input. The input histogram data can be visualized on PDF plot as points, 2-dimensional histogram or kernel density estimation (KDE) contour plot. Model mixtures are presented as lines. Additionally, components of a mixture can be plotted, both separately and in stacked mode. The tool automatically normalizes data points in the case of logarithmically-binned histograms. Moreover, the framework contains a custom fast Fourier transform (FFT) based implementation of weighted KDE computation.