tgraph 1.0.1

Implementation of the t-graph package presented in the paper 'CallMine: Fraud Detection and Visualization of Million-Scale Call Graphs'.

Tgraph: A static and temporal graph analysis tool

This is an implementation/publishing of the tgraph module used in the paper 'CallMine: Fraud Detection and Visualization of Million-Scale Call Graphs'.

For more information see Callmine: Fraud Detection and Visualizaion of Million-Scale Call Graphs.

Installation

You can install this package from PyPI:

pip install tgraph

Usage

The package offers three main feature sets: Creating static/temporal graphs from edgelists (and outputting their features), generating useful plots from features, and automatically joining CSV files resulting from graph features.

Creating static/temporal graphs from edgelists

Creating a static graph

In order to create a static graph, the user must provide a CSV file containing the following expected headers:

• "source": The source node.
• "destination": The destination node.
• "measure": A value expressing the measure of connection from source to destination.

This can either be done through the terminal (in which case tgraph performs a simple print and outputs the features if the flag -v is set):

staticgraph path/to/my/file

# or

staticgraph -v path/to/my/file

Or inside a python script:

from tgraph import static_graph

my_graph = static_graph.StaticGraph("path/to/my/file")

# Will perform a print of the adjacency dataframe
my_graph.my_print()

# Will print the graph features as csv
my_graph.print_to_csv("/path/to/output/file")
# ...

Creating a temporal graph

In order to create a static graph, the user must provide a CSV file containing the following expected headers:

• "source": The source node.
• "destination": The destination node.
• "measure": A value expressing the measure of connection from source to destination.
• "timestamp": A value expressing the timestamp associated with this connection.

This can either be done through the terminal (in which case tgraph performs a simple print and outputs the features if the flag -v is set):

temporalgraph path/to/my/file

# or

temporalgraph -v path/to/my/file

Or inside a python script:

from tgraph import temporal_graph

my_graph = temporal_graph.TemporalGraph("path/to/my/file")

# Will perform a print of the adjacency dataframe
my_graph.my_print()

# Will print the graph features as csv
my_graph.print_to_csv("/path/to/output/file")
# ...

Generating plots from graph features

The nd_cloud class provides a number of automatically generated visualizations and plots from graph feature datasets. The user is encouraged to check the source-code from the github page in order to find their desired functionality.

The module can be called from the terminal using:

ndcloud path/to/my/file

With the following optional flags:

• -v | --verbose: Enables output verbosity. Default = False
• -m | --min_row_sum: Drop rows with smaller sum. Default = 0
• -p | --print: Print to png. Default = False

It can also be used inside a Python script:

from tgraph import nd_cloud

ndc = nd_cloud.nd_cloud("path/to/my/file")

# Will simply print the headers
ndc.print_headers()

# Will produce ALL the plots
# You can set verbose to 0, print_flag to False and min_row_sum to 0 to get default behavior
ndc.kitchen_sink(verbose, print_flag, min_row_sum)

# Check out the nd_cloud class in the github repository for additional methods

Joining feature files

In order to join CSV feature files generated by the static and temporal graphs, the user can call the module from the terminal (with the optional -v flag):

joinfeaturefiles /path/to/static/file/ /path/to/temporal/file

# or

joinfeaturefiles -v /path/to/static/file/ /path/to/temporal/file

Alternatively, the module can also be used inside a Python script:

from tgraph import join_feature_files

jf = join_feature_files.JoinFeatures("path/to/static/file", "path/to/temporal/file")

jf.print_to_csv("path/to/output/file")