Authours:

Alessandro Rizzuto 187156

Francesco Contaldo 190626

Project:

Big Data Project "Frequent Pattern Mining on a single Graph"

The whole work has been done exploiting the work presented in this paper: Mohammed Elseidy, Ehab Abdelhamid, Spiros Skiadopoulos, and Panos Kalnis. "GRAMI: Frequent Subgraph and Pattern Mining in a Single Large Graph. PVLDB, 7(7):517-528, 2014."

The core algorithms and concepts have been implemented from scratch for this university project

Both input and output respect the dot notation to describe a graph https://en.wikipedia.org/wiki/DOT_(graph_description_language)

to transform a dot file into visible pdf format use 'dot -Tps NameFile.dot -o NameFile.pdf'

To generate a random input graph use the following command: 'python graphGen.py NumberOfNodes NumberOfLabel NumberOfHours(weight)'

the second parameter is referred to the number of possible values that can be assigned to a single node. Range of value valid 0..13

the third parameter is used to decide the number of possible values that the different weights (hours) can take. Range of valid value 0..24

The automated generated graph is created in the data folder with the name 'graphGenOut.dot'

To mine the graph use: spark-submit --class "App" --conf 'spark.driver.extraJavaOptions=-Xss1g' --conf 'spark.executor.extraJavaOptions=-Xss1g' JarFileName.jar

Optional --master local[n] to set the number of local nodes

Then will be asked interactively the support threshold and the subgraph maximum size

The found subgraphs are stored in the Result directory with the dot notation.