Exact discovery of the top-k sequential patterns under Leverage
This code was used to generate the results of the paper entitled: Skopus: Mining top-k sequential patterns under Leverage. When using this repository, please cite:
@article{Petitjean2016-Skopus,
author = {Petitjean, Francois and Li, Tao and Tatti, Nikolaj and Webb, Geoffrey I.},
title = {Skopus: Mining top-k sequential patterns under Leverage},
journal = {Data Mining and Knowledge Discovery},
volume = {30},
number = {5},
pages = {1086--1111},
year = {2016}
}
Note the paper is available here
The code in src is provided under GLP License version 3.
Datasets are provided for reproducibility and without any license.
Skopus requires Java 8; easiest way to build it is to import it in Eclipse. Manually, you'll have to do something like:
cd Skopus
mkdir bin
javac -d bin/ src/**/*
Then, from the binary directory, simply call
cd bin
java -Xmx1g skopus.InterestingnessMeasureMain <dataset> <output-folder> -i2 -m1 -k100
The parameters are:
-i<number>: the interestingness measure (1 for support, 2 for leverage [to be used by default])
-m<number>: m-value for the smoothing of support (1 is Lapace [to be used by default], 0 is MLE)
-k<number>: top-"how-many" patterns you want to extract
-l<number>: maximum length of the patterns
If you want have a large dataset and/or want to speed up the computation, you can limit the size to, say, patterns of length 5:
java -Xmx1g skopus.InterestingnessMeasureMain <dataset> <output> -i2 -m1 -k100 -l5
Example datasets are given in the data subfolder. Basically, each line contains one sequence, and the items are separated by spaces. Items can only be given in terms of letters or numbers (and other characters will be considered as separators).