def stase_metrics(df):
"""Compute STASE metrics for a given dataset"""
M, N = df.shape
logging.info("FILES: {0}, ANTIVIRUS: {1}".format(M, N))
clusters = df.stack().value_counts()
logging.info("clustering DONE")
O = clusters.size
logging.info("LABELS: {0}".format(O))
apps = rows_stats(df)
logging.info("rowstats computation DONE")
avs = cols_stats(df)
logging.info("colstats computation DONE")
m = pd.Series()
m['equiponderance'], m['equiponderance_idx'] = ouroboros(avs['positives'], True)
m['exclusivity'] = avs['alones'].sum() / M
m['recognition'] = apps['positives'].mean() / N
m['synchronicity'] = np.mean(avs['overlap'])
m['genericity'] = 1 - (O-1) / (avs['positives'].sum()-1)
m['uniformity'], m['uniformity_idx'] = ouroboros(clusters, True)
m['divergence'] = (apps['distincts'].sum() - M) / (apps['positives'].sum() - M)
m['consensuality'] = (apps['max'].sum() - M) / (apps['positives'].sum() - M)
m['resemblance'] = np.mean(apps['resemblance'].dropna())
m['avs'] = N
m['apps'] = M
m['labels'] = O
logging.info("metrics computation DONE")
return m, apps, avs, clusters
def test_against_gini(self):
print("{0:<15} || {1:<s} | {2:<4s}".format("ARRAY", "OURO", "GINI"))
for array, _, _ in self.cases:
orb = ouroboros(array)
gin = self._gini(array)
arr = ','.join([str(a) for a in array])
print("{0:<15} || {1:<4.2f} | {2:<4.2f}".format(arr, orb, gin))
def test_standalone(self):
for array, index, indice in self.cases:
idx, idc = ouroboros(array, True)
self.assertAlmostEqual(idx, index, 2, msg="Array: {0}".format(array))
self.assertEqual(idc, indice, msg="Array: {0}".format(array))
