def testFindAllSimilarityFromNodeOnPathSimExampleThree(self):
"""
Tests similarity for all other nodes given a single node, using example 3 from PathSim paper
"""
graph, authorMap, conferenceMap = SampleGraphUtility.constructPathSimExampleThree()
metaPath = [Author, Paper, Conference, Paper, Author]
strategy = PathSimStrategy(graph, metaPath)
mike = authorMap['Mike']
mostSimilarNodes = strategy.findMostSimilarNodes(mike, 5)
self.assertEquals([authorMap['Bob'], authorMap['Mary'], authorMap['Jim']], mostSimilarNodes)
def testFindAllSimilarityFromNodeOnPathSimExampleThree(self):
"""
Tests similarity for all other nodes given a single node, using example 3 from PathSim paper
"""
graph, authorMap, conferenceMap = SampleGraphUtility.constructPathSimExampleThree(
)
metaPath = [Author, Paper, Conference, Paper, Author]
strategy = PathSimStrategy(graph, metaPath)
mike = authorMap['Mike']
mostSimilarNodes = strategy.findMostSimilarNodes(mike, 5)
self.assertEquals(
[authorMap['Bob'], authorMap['Mary'], authorMap['Jim']],
mostSimilarNodes)
def run(self):
strategy = PathSimStrategy(
self.graph, [Conference, Paper, Author, Paper, Conference], True)
experimentHelper = LabeledExperimentHelper(
os.path.join('data', 'dbis', 'query_label', 'PathSim'))
conferenceQueryNames = [
'SIGMOD Conference', 'VLDB', 'ICDE', 'PODS', 'EDBT', 'DASFAA',
'KDD', 'ICDM', 'PKDD', 'SDM', 'PAKDD', 'WWW', 'SIGIR', 'TREC',
'APWeb'
]
for conferenceQueryName in conferenceQueryNames:
conferences = experimentHelper.getNodesByAttribute(
self.graph, 'name', conferenceQueryName)
assert (len(conferences) == 1)
target = list(conferences)[0]
number = 10
# Output the top ten most similar conferences on the CPAPC meta path
self.output(
'\n\nTop Ten Similar Conferences to %s (CPAPC meta path):' %
conferenceQueryName)
mostSimilarNodes = strategy.findMostSimilarNodes(target, number)
apaPathTable = texttable.Texttable()
headerRow = [['Rank', 'Conference', 'Relevance']]
dataRows = [[
i + 1, mostSimilarNodes[i].name,
experimentHelper.getLabelForNode(target, mostSimilarNodes[i])
] for i in xrange(0, number)]
apaPathTable.add_rows(headerRow + dataRows)
self.output(apaPathTable.draw())
# Output the nDCG for these results
self.output(
'%1.3f' %
CumulativeGainMeasures.normalizedDiscountedCumulativeGain(
target, mostSimilarNodes,
experimentHelper.labelDictionary))
def run(self):
strategy = PathSimStrategy(self.graph, [Conference, Paper, Author, Paper, Conference], True)
experimentHelper = LabeledExperimentHelper(os.path.join('data', 'dbis', 'query_label', 'PathSim'))
conferenceQueryNames = [
'SIGMOD Conference',
'VLDB',
'ICDE',
'PODS',
'EDBT',
'DASFAA',
'KDD',
'ICDM',
'PKDD',
'SDM',
'PAKDD',
'WWW',
'SIGIR',
'TREC',
'APWeb'
]
for conferenceQueryName in conferenceQueryNames:
conferences = experimentHelper.getNodesByAttribute(self.graph, 'name', conferenceQueryName)
assert(len(conferences) == 1)
target = list(conferences)[0]
number = 10
# Output the top ten most similar conferences on the CPAPC meta path
self.output('\n\nTop Ten Similar Conferences to %s (CPAPC meta path):' % conferenceQueryName)
mostSimilarNodes = strategy.findMostSimilarNodes(target, number)
apaPathTable = texttable.Texttable()
headerRow = [['Rank', 'Conference', 'Relevance']]
dataRows = [[i + 1, mostSimilarNodes[i].name, experimentHelper.getLabelForNode(target, mostSimilarNodes[i])] for i in xrange(0, number)]
apaPathTable.add_rows(headerRow + dataRows)
self.output(apaPathTable.draw())
# Output the nDCG for these results
self.output('%1.3f' % CumulativeGainMeasures.normalizedDiscountedCumulativeGain(target, mostSimilarNodes, experimentHelper.labelDictionary))
