def calculateAverageConceptVectorDistance(self, concepts, filterName='DSM'):
'''
This function calculates the average concept vector distance, based on the vector representing a cui.
'''
feats = dict()
if (filterName=='ALL'):
feats.update(self.calculateAverageConceptVectorDistance(concepts, 'DSM'))
feats.update(self.calculateAverageConceptVectorDistance(concepts, 'DSM+1'))
feats.update(self.calculateAverageConceptVectorDistance(concepts, 'MED'))
else:
subset = list(set(self.getSubsetOfConcepts(concepts, filterName)))
maxDist = 0
dists = []
for cui in subset:
for cui2 in subset:
try:
concept = cui.split(';')[0]
wVector= Resources.getConceptVectors_dsm().vectorize(concept, remove_oov=True)
concept2 = cui2.split(';')[0]
wVector2= Resources.getConceptVectors_dsm().vectorize(concept2, remove_oov=True)
cosDist = spatial.distance.cosine(wVector, wVector2)
if not math.isnan(cosDist):
dists.append(cosDist)
if (cosDist) > maxDist:
maxDist = cosDist
except:
continue
feats[filterName+'maxConceptVectorDist'] = round(maxDist,3)
if bool(dists):
feats[filterName+'avgConceptVectorDist'] = round(np.mean(dists),3)
return feats
def getVectorizedConcepts(self, concepts):
"""
Vectorizes a list of Concept Unique Identifiers into an array of concept vectors of identical size.
@param concepts: A list of CUIS.
@return: A list of vectors, each one of which represents a concept.
"""
return Resources.getConceptVectors_dsm().vectorize(concepts, remove_oov=False)
