Failure Clustering

Environment

• Developed & tested under Python 3.9.1
• Installing dependencies:

python -m pip install -r requirements.txt

Installation

git clone https://github.com/Suresoft-GLaDOS/failure_clustering
cd failure_clustering
pip install setuptools
python setup.py install

Getting started

! The full example script is provided in main.ipynb

Calculating the distance between failing test cases via hypergraph modeling

• prerequisite: SBFL-engine

from sbfl.base import SBFL
from failure_clustering.base import FailureDistance

test_names = ["T1", "T2", "T3", "T4", "T5"]

# Coverage of T1, ..., T5
X = [
    [0, 1, 1, 0, 1, 0], # Coverage of T1 
    [1, 0, 0, 1, 0, 0], # Coverage of T2
    [1, 1, 0, 0, 1, 1], # Coverage of T3
    [0, 1, 0, 1, 1, 0], # Coverage of T4
    [1, 1, 0, 0, 1, 1], # Coverage of T5
]

# Test results of T1, ..., T5
y = [0, 0, 1, 0, 1] # 0: FAIL, 1: PASS

#Calculating weights of program elements
sbfl = SBFL(formula='Tarantula')
w = sbfl.fit_predict(X, y)
print(w)
"""
[0.25 0.4  1.   1.   0.4  0.  ]
"""

fd = FailureDistance(measure='hdist')
distance_matrix, failure_indices = fd.get_distance_matrix(
    X, y, weights=w, return_index=True)

# pairwise distances among failing test cases (T1, T2, T4)
print(distance_matrix) 
"""
[[0.         1.         0.77380952]
 [1.         0.         0.21428572]
 [0.77380952 0.21428572 0.        ]]
"""
print(failure_indices)
"""
[0 1 3]
"""

• Supported measures for FailureDistance
  • jaccard, braycurtis, canberra, chebyshev, cityblock, correlation, cosine, dice, euclidean, hamming, jaccard, jensenshannon, kulsinski, kulczynski1, mahalanobis, matching, minkowski, rogerstanimoto, russellrao, seuclidean, sokalmichener, sokalsneath, sqeuclidean, yule, hdist

Running Agglomerative Hierarchical Clustering

You can provide a stopping criterion when performing clustering. If the stopping criterion is not provided, a list of clustering results at all iterations are returned.

1. The stopping criterion can be a threshold value, such as 0.5. The clustering stops once the minimum intercluster distance exceeds the threshold.

   from failure_clustering.clustering import Agglomerative
   
   aggl = Agglomerative(linkage='complete')
   clustering = aggl.run(distance_matrix, stopping_criterion=0.5)
   
   print(clustering)
   for i, cluster in zip(failure_indices, clustering):
       print(f"Cluster of {test_names[i]}: {cluster}")
   """
   [0, 1, 1]
   Cluster of T1: 0
   Cluster of T2: 1
   Cluster of T4: 1
   """

   • FYI, aggl.mdist stores the minimum intercluster distance at each iteration.

2. min_intercluster_distance_elbow stops merging clusters at the elbow point of the minimum intercluster distance curve

   from failure_clustering.clustering import Agglomerative
   
   aggl = Agglomerative(linkage='complete')
   clustering = aggl.run(distance_matrix, 
       stopping_criterion='min_intercluster_distance_elbow')
   
   print(clustering)
   for i, cluster in zip(failure_indices, clustering):
       print(f"Cluster of {test_names[i]}: {cluster}")
   """
   [0, 1, 1]
   Cluster of T1: 0
   Cluster of T2: 1
   Cluster of T4: 1
   """