def getPrecisionRecallFalseRate(self, resultSimulation, kMAX, plot=False, output='dict'):
"""
The function receives the detections/false alarms of all the nodes
and computes the Precision/Recall for all the K.
"""
precisionConfInterval = {}
recallConfInterval = {}
falseConfInterval = {}
for k in range(1, kMAX + 1):
precisionConfInterval[k] = []
recallConfInterval[k] = []
falseConfInterval[k] = []
for key, value in resultSimulation.iteritems():
detections = value['detections']
falsePositives = value['falsePositives']
for k in range(1, kMAX + 1):
if detections[k]['events'] != 0:
recallConfInterval[k].append(detections[k]['detection']/float(detections[k]['events']))
if (detections[k]['detection'])!= 0 and falsePositives != 0:
precisionConfInterval[k].append((detections[k]['detection'])/(float(detections[k]['detection']) + falsePositives[k] ))
falseConfInterval[k].append(falsePositives[k]/float(len(self.truth.clears)))
errorRecall = np.ndarray(kMAX)
errorPrecision = np.ndarray(kMAX)
errorFalse = np.ndarray(kMAX)
meanRecall = np.ndarray(kMAX)
meanPrecision = np.ndarray(kMAX)
meanFalseRate = np.ndarray(kMAX)
for k in range(1, kMAX + 1):
a = recallConfInterval[k]
meanRecall[k-1] = np.mean(a)
interval = sms.DescrStatsW(a).tconfint_mean()
errorRecall[k-1] = interval[1] - np.mean(a)
a = precisionConfInterval[k]
meanPrecision[k-1] = np.mean(a)
interval = sms.DescrStatsW(a).tconfint_mean()
errorPrecision[k-1] = interval[1] - np.mean(a)
a = falseConfInterval[k]
meanFalseRate[k-1] = np.mean(a)
interval = sms.DescrStatsW(a).tconfint_mean()
errorFalse[k-1] = interval[1] - np.mean(a)
if plot:
visual = Visualization()
visual.barRecallPrecisionvsK2(meanRecall, meanFalseRate, meanPrecision, errorRecall, errorPrecision, errorFalse)
if output == 'dict':
if meanFalseRate[0] > 1:
meanFalseRate[0] = 1
result = {
'Precision': np.nan_to_num(meanPrecision).tolist(),
'errPrecision': errorPrecision.tolist(),
'Recall': np.nan_to_num(meanRecall).tolist(),
'errRecall': errorRecall.tolist(),
'FalseRate': np.nan_to_num(meanFalseRate).tolist(),
'errFalseRate': errorFalse.tolist()
}
return result
elif output == 'tuple':
return meanPrecision, errorPrecision, meanRecall, errorRecall, meanFalseRate, errorFalse
else:
return
