def megaVector(ittRec, bc_dIDF, bUserStopFilter, bc_lStopWords, nTot):
dThisPoint = {}
dUsableWords = bc_dIDF.value
norm = 0.0
for post in ittRec:
for w in fspLib.wordBreak(post.text, bUserStopFilter, bc_lStopWords):
if w in dUsableWords:
if dUsableWords[w][0] != 0:
val = log10((1.*nTot)/dUsableWords[w][0])
norm = norm + val
dThisPoint[w] = dThisPoint[w] + val if w in dThisPoint.keys() else val
scoreNPos = []
if norm == 0.0:
return SparseVector(len(dUsableWords.keys()),[],[])
for term, num in dThisPoint.iteritems():
scoreNPos.append((dUsableWords[term][0],num/norm))
scoreNPos.sort(key=lambda x:x[0])
lPos, lVal = zip(*scoreNPos)
return SparseVector(len(dUsableWords.keys()), list(lPos), list(lVal))
def megaVector(ittRec, bc_dIDF, bUserStopFilter, bc_lStopWords, nTot):
dThisPoint = {}
dUsableWords = bc_dIDF.value
norm = 0.0
for post in ittRec:
for w in fspLib.wordBreak(post.text, bUserStopFilter, bc_lStopWords):
if w in dUsableWords:
if dUsableWords[w][0] != 0:
val = log10((1. * nTot) / dUsableWords[w][0])
norm = norm + val
dThisPoint[w] = dThisPoint[
w] + val if w in dThisPoint.keys() else val
scoreNPos = []
if norm == 0.0:
return SparseVector(len(dUsableWords.keys()), [], [])
for term, num in dThisPoint.iteritems():
scoreNPos.append((dUsableWords[term][0], num / norm))
scoreNPos.sort(key=lambda x: x[0])
lPos, lVal = zip(*scoreNPos)
return SparseVector(len(dUsableWords.keys()), list(lPos), list(lVal))
def locationBasedOutput(bByDate, jobNm, vecAndPredictions, sNum, fBinSize, revLookup, bUseStopFilter, bc_lStopWords):
nNonZeros = vecAndPredictions.count()
nToTake = int(nNonZeros*sNum)
if sNum>1.:
nToTake=int(sNum)
outCol = vecAndPredictions.sortBy(lambda x: x[1], False).take(nToTake)
if bByDate == True:
datesJson = {}
for point in outCol:
record = point[0][1]
breakPoint = record[0].find("_",9)
offset = fBinSize/2.
lat = float(record[0][9:breakPoint])
if lat > 0:
lat = lat + offset
else:
lat = lat - offset
lon = float(record[0][breakPoint+1:])
if lon > 0:
lon = lon + offset
else:
lon = lon - offset
sdate = record[0][:8]
sdate = sdate[0:4]+"-"+sdate[4:6]+"-"+sdate[6:]
if sdate not in datesJson.keys():
datesJson[sdate] = {"clusters":[]}
thisCluster = {"nUnique":5,"background":100, "nTotal":len(record[1]), "lon":lon, "lat":lat, "date":sdate, "posts":[], "score":point[1]}
labeledP = point[0][0]
tups = zip(labeledP.features.values, labeledP.features.indices)
thisDict = set(map(lambda x: revLookup[x[1]], sorted(tups, key=lambda x: x[0], reverse=True)[:5]))
thisCluster["dict"] = list(thisDict)
for post in record[1]:
includePost = False
for w in fspLib.wordBreak(post.text, bUseStopFilter, bc_lStopWords):
if w in thisDict:
includePost = True
break
if includePost:
thisPost = {"sco":1,"cap":post.text,"lon":post.lon,"lat":post.lat,"date":post.dt.strftime("%Y-%m-%d"),"usr":post.user,"source":post.source, "datetime": post.dt.strftime("%Y-%m-%d %H:%M:%S")}
thisCluster["posts"].append(thisPost)
thisCluster["poly"] = [[lat+offset,lon+offset],[lat+offset,lon-offset],[lat-offset,lon-offset],[lat-offset,lon+offset]]
datesJson[sdate]["clusters"].append(thisCluster)
retDict = {"type":"event", "dates":datesJson}
with codecs.open("scoreFiles/"+jobNm, encoding="utf-8",mode="wb") as fOut:
json.dump(retDict, fOut)
return retDict
else:
clusterList = []
for point in outCol:
record = point[0][1]
breakPoint = record[0].find("_")
offset = fBinSize/2.
lat = float(record[0][:breakPoint])
if lat > 0:
lat = lat + offset
else:
lat = lat - offset
lon = float(record[0][breakPoint+1:])
if lon > 0:
lon = lon + offset
else:
lon = lon - offset
thisCluster = {"nUnique":5,"background":100, "nTotal":len(record[1]), "lon":lon, "lat":lat, "posts":[], "score":point[1]}
labeledP = point[0][0]
tups = zip(labeledP.features.values, labeledP.features.indices)
thisDict = set(map(lambda x: revLookup[x[1]], sorted(tups, key=lambda x: x[0], reverse=True)[:5]))
thisCluster["dict"] = list(thisDict)
nPosts = 0
for post in record[1]:
includePost = False
for w in fspLib.wordBreak(post.text, bUseStopFilter, bc_lStopWords):
if w in thisDict:
includePost = True
break
if includePost:
nPosts = nPosts + 1
thisPost = {"sco":1,"cap":post.text,"lon":post.lon,"lat":post.lat,"date":post.dt.strftime("%Y-%m-%d"),"usr":post.user,"source":post.source, "datetime": post.dt.strftime("%Y-%m-%d %H:%M:%S")}
thisCluster["posts"].append(thisPost)
if nPosts >= 100:
break
thisCluster["poly"] = [[lat+offset,lon+offset],[lat+offset,lon-offset],[lat-offset,lon-offset],[lat-offset,lon+offset]]
clusterList.append(thisCluster)
retDict = {"type":"place", "clusters":clusterList}
with codecs.open("scoreFiles/"+jobNm, encoding="utf-8",mode="wb") as fOut:
json.dump(retDict, fOut)
return retDict
nDataType = 1
if args.datTyp:
nDataType = args.datTyp
cNum = 1200
if args.cNum:
cNum = args.cNum
bUseStopFilter = False
#Declare Spark Context
conf = SparkConf().setAppName(jobNm)
sc = SparkContext(conf = conf)
lowerTime = date(2006,03,21)
upperTime = date(3000,01,01)
bc_lStopWords = None
input = sc.textFile(inputFile,100).coalesce(cNum)
data = input.map(lambda x: fspLib.loadRecord(x,nDataType)).filter(lambda x: x is not None)
df0 = data.filter(lambda x: fspLib.badData(x, bUseStopFilter, bc_lStopWords, lowerTime, upperTime))
wb = df0.flatMap(lambda x: fspLib.wordBreak(x['caption'], bUseStopFilter, bc_lStopWords))
wc = wb.map(lambda x: (x,1)).reduceByKey(add)
wordAndCound = wc.collect()
fDict = codecs.open("./idfFiles/" + dictFile, encoding="utf-8", mode="w")
for cT in wordAndCound:
fDict.write(u"\t".join([unicode(cT[0]),unicode(cT[1])])+u"\n")
fDict.close()
if args.datTyp:
nDataType = args.datTyp
cNum = 1200
if args.cNum:
cNum = args.cNum
bUseStopFilter = False
#Declare Spark Context
conf = SparkConf().setAppName(jobNm)
sc = SparkContext(conf=conf)
lowerTime = date(2006, 03, 21)
upperTime = date(3000, 01, 01)
bc_lStopWords = None
input = sc.textFile(inputFile, 100).coalesce(cNum)
data = input.map(lambda x: fspLib.loadRecord(x, nDataType)).filter(
lambda x: x is not None)
df0 = data.filter(lambda x: fspLib.badData(
x, bUseStopFilter, bc_lStopWords, lowerTime, upperTime))
wb = df0.flatMap(lambda x: fspLib.wordBreak(x['caption'], bUseStopFilter,
bc_lStopWords))
wc = wb.map(lambda x: (x, 1)).reduceByKey(add)
wordAndCound = wc.collect()
fDict = codecs.open("./idfFiles/" + dictFile, encoding="utf-8", mode="w")
for cT in wordAndCound:
fDict.write(u"\t".join([unicode(cT[0]), unicode(cT[1])]) + u"\n")
fDict.close()
def locationBasedOutput(bByDate, jobNm, vecAndPredictions, sNum, fBinSize,
revLookup, bUseStopFilter, bc_lStopWords):
nNonZeros = vecAndPredictions.count()
nToTake = int(nNonZeros * sNum)
if sNum > 1.:
nToTake = int(sNum)
outCol = vecAndPredictions.sortBy(lambda x: x[1], False).take(nToTake)
if bByDate == True:
datesJson = {}
for point in outCol:
record = point[0][1]
breakPoint = record[0].find("_", 9)
offset = fBinSize / 2.
lat = float(record[0][9:breakPoint])
if lat > 0:
lat = lat + offset
else:
lat = lat - offset
lon = float(record[0][breakPoint + 1:])
if lon > 0:
lon = lon + offset
else:
lon = lon - offset
sdate = record[0][:8]
sdate = sdate[0:4] + "-" + sdate[4:6] + "-" + sdate[6:]
if sdate not in datesJson.keys():
datesJson[sdate] = {"clusters": []}
thisCluster = {
"nUnique": 5,
"background": 100,
"nTotal": len(record[1]),
"lon": lon,
"lat": lat,
"date": sdate,
"posts": [],
"score": point[1]
}
labeledP = point[0][0]
tups = zip(labeledP.features.values, labeledP.features.indices)
thisDict = set(
map(lambda x: revLookup[x[1]],
sorted(tups, key=lambda x: x[0], reverse=True)[:5]))
thisCluster["dict"] = list(thisDict)
for post in record[1]:
includePost = False
for w in fspLib.wordBreak(post.text, bUseStopFilter,
bc_lStopWords):
if w in thisDict:
includePost = True
break
if includePost:
thisPost = {
"sco": 1,
"cap": post.text,
"lon": post.lon,
"lat": post.lat,
"date": post.dt.strftime("%Y-%m-%d"),
"usr": post.user,
"source": post.source,
"datetime": post.dt.strftime("%Y-%m-%d %H:%M:%S")
}
thisCluster["posts"].append(thisPost)
thisCluster["poly"] = [[lat + offset, lon + offset],
[lat + offset, lon - offset],
[lat - offset, lon - offset],
[lat - offset, lon + offset]]
datesJson[sdate]["clusters"].append(thisCluster)
retDict = {"type": "event", "dates": datesJson}
with codecs.open("scoreFiles/" + jobNm, encoding="utf-8",
mode="wb") as fOut:
json.dump(retDict, fOut)
return retDict
else:
clusterList = []
for point in outCol:
record = point[0][1]
breakPoint = record[0].find("_")
offset = fBinSize / 2.
lat = float(record[0][:breakPoint])
if lat > 0:
lat = lat + offset
else:
lat = lat - offset
lon = float(record[0][breakPoint + 1:])
if lon > 0:
lon = lon + offset
else:
lon = lon - offset
thisCluster = {
"nUnique": 5,
"background": 100,
"nTotal": len(record[1]),
"lon": lon,
"lat": lat,
"posts": [],
"score": point[1]
}
labeledP = point[0][0]
tups = zip(labeledP.features.values, labeledP.features.indices)
thisDict = set(
map(lambda x: revLookup[x[1]],
sorted(tups, key=lambda x: x[0], reverse=True)[:5]))
thisCluster["dict"] = list(thisDict)
nPosts = 0
for post in record[1]:
includePost = False
for w in fspLib.wordBreak(post.text, bUseStopFilter,
bc_lStopWords):
if w in thisDict:
includePost = True
break
if includePost:
nPosts = nPosts + 1
thisPost = {
"sco": 1,
"cap": post.text,
"lon": post.lon,
"lat": post.lat,
"date": post.dt.strftime("%Y-%m-%d"),
"usr": post.user,
"source": post.source,
"datetime": post.dt.strftime("%Y-%m-%d %H:%M:%S")
}
thisCluster["posts"].append(thisPost)
if nPosts >= 100:
break
thisCluster["poly"] = [[lat + offset, lon + offset],
[lat + offset, lon - offset],
[lat - offset, lon - offset],
[lat - offset, lon + offset]]
clusterList.append(thisCluster)
retDict = {"type": "place", "clusters": clusterList}
with codecs.open("scoreFiles/" + jobNm, encoding="utf-8",
mode="wb") as fOut:
json.dump(retDict, fOut)
return retDict
