def on_data(self, data):
tweet = GetTweets.get(data)
if not tweet:
return
tag = self.host.tags[0]
indx = 0
for t in self.host.tags:
if t in tweet:
tag = t
break
indx += 1
indx = min(indx, len(self.host.tags)-1)
keywords = self.host.getKeywords.get(tweet, tag)
if not keywords['words']:
return
if tag == self.host.correct_tag:
keywords["label"] = 1
else:
keywords["label"] = 0
try:
self.host.cache[indx] = keywords
except:
pass
def on_data(self, data):
tweet = GetTweets.get(data)
if not tweet:
return
print tweet
keywords = GetKeywords.get(tweet, sys.argv[1])
if not keywords['words']:
return
def on_data(self, data):
tweet = GetTweets.get(data)
if not tweet:
return
keywords = GetKeywords.get(tweet, sys.argv[1])
if not keywords["words"]:
return
print keywords
# winnow
self.winnow.add(keywords['words'])
label = self.winnow.predict(keywords['words'])
self.winnow.learn(label, keywords)
# Evalutation
fact = keywords['label']
self.fp *= c
self.tp *= c
self.fn *= c
self.tn *= c
if label > fact:
self.fp += 1
elif label < fact:
self.tn += 1
elif label == 1:
self.tp += 1
else:
self.fn += 1
acc = (self.tp + self.fn)/(self.tp + self.fn + self.tn + self.fp + 1)
pre = self.tp / (self.tp + self.fp + 1)
rec = self.tp / (self.tp + self.fn + 1)
acc = int(acc * 100) / 100.0
pre = int(pre * 100) / 100.0
rec = int(rec * 100) / 100.0
print ">>>> Get %s but actually was %s" % (label, fact)
print ">>>> Accuracy: %s,\t Precision: %s,\t Recall: %s" % (acc, pre, rec)
print ">>>> List length: %s" % len(self.winnow.stack)
