def scatterPlot(clustering, location, fileName):
userClusterMap = {}
for clusterId, users in clustering[2]['clusters'].iteritems():
for user in users:
if user in location['users']: userClusterMap[user]=clusterId
scatterData = defaultdict(dict)
clusterMap = clustering[3]
for user, userVector in location['users'].iteritems():
if user in userClusterMap:
for d in userVector:
for db in userVector[d]:
for h in [(datetime.datetime.fromtimestamp(ep).hour-6)%24 for ep in userVector[d][db]]:
if h not in scatterData[userClusterMap[user]]: scatterData[userClusterMap[user]][h]=0
scatterData[userClusterMap[user]][h]+=1
total = float(sum([k for cluster, clusterInfo in scatterData.iteritems() for k, v in clusterInfo.iteritems() for i in range(v)]))
for cluster, clusterInfo in scatterData.iteritems():
if cluster in validClusters:
# if type=='normal':
data = [k for k, v in clusterInfo.iteritems() for i in range(v)]
mean, std = np.mean(data), np.std(data)
if std!=0: plotNorm(sum(data)/total, mean, std, color=clusterMap[cluster])
else: plotNorm(sum(data)/total, mean, random.uniform(0.1, 0.5), color=clusterMap[cluster])
# elif type=='scatter': plt.scatter(clusterInfo.keys(), clusterInfo.values(), color=clusterMap[cluster], label=cluster)
plt.title('%s (%s, %s, %s)'%(location['name'],location['location'], location['categories'], location['tags'])),plt.legend()
# plt.show()
plt.xlim(xmin=0,xmax=24)
plt.savefig(fileName), plt.clf()
def plotLocation(locationName, locationId, locationClustering, dayBlockMeans, dayBlockStandardDeviations, colorMap):
classes, classDistribution = getDataDistribution(locationClustering.values())
mu, sigma = dayBlockMeans, dayBlockStandardDeviations
totalUsers = float(sum(classDistribution))
for dist, mu, sigma, color in zip(classDistribution, mu, sigma, [colorMap[c] for c in classes]):
if sigma==0: sigma=0.15
plotNorm(dist/totalUsers, scale(mu), scale(sigma), color=color)
plt.title('%s (%s)'%(locationName,locationId))
plt.xlim(xmin=0,xmax=24)
# plt.show()
plt.savefig(fileName)
plt.clf()
def plotGaussianGraphsForClusters(place):
for location in Analysis.iterateLocationsWithClusterDetails(place):
total = location['total']
clustersInfo = location['clustersInfo']
for clusterId, data in clustersInfo.iteritems():
mean, std, clusterSum, color = data['mean'], data['std'], data['clusterSum'], data['color']
if std!=0: plotNorm(clusterSum/total, mean, std, color=color, label=str(clusterId))
else: plotNorm(clusterSum/total, mean, random.uniform(0.1, 0.5), color=color, label=str(clusterId))
plt.xlim(xmin=0, xmax=23); plt.legend()
plt.title(location['name'])
fileName = '/'.join([placesGaussianImagesFolder%place['name'], getLocationType(location), location['location'].replace(' ', '_').replace('.', '+')+'.png'])
print fileName
FileIO.createDirectoryForFile(fileName)
plt.savefig(fileName), plt.clf()
def plotLocationDistribution():
'''Types of locations seen:
=> Locations where different people have to be at same time: Example office, pub
=> Locations that different people choose to go at different times: cafe+party place
Big cluster suggests most people who come to a location go to similar locations (implies similar people).
Their mean suggests the most poplar time to go to that location.
'''
def scale(val): return (val*4)+2#val*2*4+2
for location in FileIO.iterateJsonFromFile(locationClustersFile):
if 'clustering' in location:
classes, classDistribution = getDataDistribution(location['clustering'][1].values())
mu, sigma = location['clustering'][2][0], location['clustering'][2][1]
totalUsers = float(sum(classDistribution))
for dist, mu, sigma in zip(classDistribution, mu, sigma):
if sigma==0: sigma=0.15
print dist/totalUsers
plotNorm(dist/totalUsers, scale(mu), scale(sigma))
title = venuesCollection.find_one({'lid':location['location']})
if title!=None: title = unicode(title['n']).encode("utf-8")
else: title = ''
plt.title('%s (%s)'%(title,location['location']))
plt.xlim(xmin=0,xmax=24)
print 'comes here'
plt.show()
