A fast implementation of the PaNDa+ algorithm for mining Top-K Binary patterns in transactional data (C. Lucchese, S. Orlando, R. Perego – A Unifying Framework for Mining Approximate Top-K Binary Patterns).
mkdir build
cd build
cmake ..
make./speedy-panda [filename] {OPTIONS}
Finds approximate patterns in datasets with noise.
OPTIONS:
-h, --help Display this help menu
-k[8] Max number of patterns
-w[0.8] Pattern complexity weight
-x[1.0] Row tollerance ratio
-y[1.0] Column tollerance ratio
filename The dataset to processPerform Top-K Binary pattern mining with 8 threads on the file accidents_stripped2.dat, looking for 8 patterns with a pattern complexity weight of 0.5, row tollerance ratio of 0.4 and column tollerance ratio of 0.4.
OMP_NUM_THREADS=8 ./speedy-panda ../dataset/accidents_stripped2.dat -w 0.5 -k 5 -x 0.4 -y 0.4
Patterns:
12 16 17 18 21 27 29 31 43 (60833)
8 14 15 22 24 25 28 41 59 (48565)
1 17 28 38 43 63 (43529)
10 17 22 25 43 307 (22971)
23 30 35 41 (25569)
________________________________________________________
Executed in 148.52 secs fish external
usr time 1159.12 secs 136.00 micros 1159.12 secs
sys time 7.57 secs 913.00 micros 7.57 secs