def main():
print "Started---------------------------"
start_time = time.time()
thresh = ["10", "20"] #["10","15", "20", "25"]
workbook = xlrd.open_workbook(SETTINGS.dataset_folder + "dataset.xlsx")
stop_words = open(SETTINGS.stop_words_file).read().splitlines()
tweets, num_rows, tweetLabel, followerDict = ut.tweet_from_source_with_Followers(
workbook, stop_words)
rumor_category = {"NA", "R1", "R3", "R4", "R5", "R7", "All"}
conceptDict = ut.readConceptDictionary()
for th in thresh:
rumor_type = {'Description'}
for t in sorted(rumor_type):
scores = {}
rumor_list = ut.get_rumor_list(workbook, t, stop_words)
rumor_word_count = {i: {} for i in range(rumor_list.__len__())}
rumor_edge_count = {i: {} for i in range(rumor_list.__len__())}
k = 0
lookedInWordnet = set()
for rumor in rumor_list:
word_array = rumor.split()
word_dict = {word: 0 for word in rumor.split()}
edge_dict = ut.tuples(word_array)
rumor_edge_count[k] = edge_dict
rumor_word_count[k] = word_dict
k += 1
threshold = np.arange(0, 0.5, 0.01)
original_rumor_word_count = rumor_word_count
original_rumor_edge_count = rumor_edge_count
word_count_storage = {}
for val in threshold:
i = 0
rumor_edge_count = original_rumor_edge_count
rumor_word_count = original_rumor_word_count
jac_labeled_tweets, rumor_word_count = ut.labelDataJaccardConcept(
tweets, rumor_word_count, val, rumor_list, conceptDict, th,
followerDict)
g_tweets = []
na_tweets = []
for key in jac_labeled_tweets:
if key == 0 or key == 2 or key == 3 or key == 4 or key == 6:
g_tweets += jac_labeled_tweets[key]
else:
na_tweets += jac_labeled_tweets[key]
labeled_tweets, rumor_edge_count = ut.labelGraphConcept(
g_tweets, rumor_edge_count, val, rumor_list.__len__(),
conceptDict, th, followerDict)
labeled_tweets[11] += na_tweets
# word_count_storage[val] = rumor_word_count
word_count_storage[val] = rumor_edge_count
workbook_new = xlwt.Workbook()
sheet = workbook_new.add_sheet("confusion Matrix")
sheet.write(0, 0, "Tweets")
sheet.write(0, 1, "Category")
sheet.write(0, 2, "True Value")
# workbook = xlrd.open_workbook("Files/"+type+"_out_"+i+".xls")
# worksheet = workbook.sheet_by_name('Sheet1')
# num_rows = worksheet.nrows - 1
tp = {}
tn = {}
fp = {}
fn = {}
excel = {}
for key in labeled_tweets:
cat = None
if key != 0 and key != 2 and key != 3 and key != 4 and key != 6:
cat = "NA"
elif key == 0:
cat = "R1"
elif key == 2:
cat = "R3"
elif key == 3:
cat = "R4"
elif key == 4:
cat = "R5"
elif key == 6:
cat == "R7"
value = labeled_tweets[key]
for item in value:
if item in tweetLabel:
type = tweetLabel[item]
if type != "NA" and cat != "NA":
if "All" not in tp:
tp["All"] = 1
else:
tp["All"] = tp["All"] + 1
elif type == "NA" and cat != "NA":
if "All" not in fp:
fp["All"] = 1
else:
fp["All"] = fp["All"] + 1
elif type != "NA" and cat == "NA":
if "All" not in fn:
fn["All"] = 1
else:
fn["All"] = fn["All"] + 1
if type == cat:
if type not in tp:
tp[type] = 1
else:
tp[type] = tp[type] + 1
else:
if cat not in fp:
fp[cat] = 1
else:
fp[cat] = fp[cat] + 1
if type not in fn:
fn[type] = 1
else:
fn[type] = fn[type] + 1
excel[item] = cat
for key in rumor_category:
tn[key] = tweets.__len__() - tp.get(key, 0) - fp.get(
key, 0) - fn.get(key, 0)
tptnlist = []
tptnlist.append(tp)
tptnlist.append(fp)
tptnlist.append(fn)
tptnlist.append(tn)
scores[val] = tptnlist
row = 1
default = 'NA'
print val
# for tweet in tweets:
# cc = excel.get(tweet)
# sheet.write(row,0,tweet.decode('ISO-8859-1'))#.decode('utf-8')
# sheet.write(row,1,cc)
# sheet.write(row,2,tweetLabel.get(tweet))
# row += 1
# save_path = os.getcwd()+"/Files/Combined/"+t
# workbook_new.save(save_path+"/out_"+str(val).encode('utf-8')+".xls")
i += 1
sorted_scores = sorted(scores.items(), key=operator.itemgetter(0))
# cPickle.dump(word_count_storage, open("word_count_storage_Graph_"+str(t)+".p", "wb"))
# cPickle.dump(scores, open("scores_Graph_"+t+".p", "wb"))
# scores = cPickle.load(open("scores_Jacc"+t+".p", "rb"))
ut.calculations(sorted_scores, rumor_category, t, th, "Combined")
workbook_eval = xlwt.Workbook()
sheet = workbook_eval.add_sheet("Evaluation")
sheet.write(0, 0, "Threshold")
j = 0
for key in rumor_category:
if key == "R7" or key == "R5" or key == "R4":
continue
sheet.write(0, j + 1, key + "_True Positive")
sheet.write(0, j + 2, key + "_False positive")
sheet.write(0, j + 3, key + "_True Negative")
sheet.write(0, j + 4, key + "_False Negative")
sheet.write(0, j + 5, key + "_Precision")
sheet.write(0, j + 6, key + "_Recall")
sheet.write(0, j + 7, key + "_F1-Measure")
sheet.write(0, j + 8, key + "_Accuracy")
j += 8
row = 1
for score, tptnlist in sorted_scores:
sheet.write(row, 0, score)
# tptnlist = sorted_scores[val];
tp1 = tptnlist[0]
fp1 = tptnlist[1]
fn1 = tptnlist[2]
tn1 = tptnlist[3]
j = 0
for key in rumor_category:
if key == "R7" or key == "R5" or key == "R4":
continue
p_1 = tp1.get(key, 0) + fp1.get(key, 0)
precision = 0
if p_1 != 0:
precision = float(tp1.get(key, 0)) / (p_1)
recall = float(tp1.get(
key, 0)) / (tp1.get(key, 0) + fn1.get(key, 0))
sheet.write(row, j + 1, tp1.get(key, 0))
sheet.write(row, j + 2, fp1.get(key, 0))
sheet.write(row, j + 3, tn1.get(key, 0))
sheet.write(row, j + 4, fn1.get(key, 0))
sheet.write(row, j + 5, precision)
sheet.write(row, j + 6, recall)
if precision + recall == 0:
sheet.write(row, j + 7, 0)
else:
sheet.write(
row, j + 7,
float(2 * precision * recall /
(precision + recall)))
accuracy = float(
(tp1.get(key, 0) + tn1.get(key, 0)) / float(
tp1.get(key, 0) + tn1.get(key, 0) +
fp1.get(key, 0) + fn1.get(key, 0)))
sheet.write(row, j + 8, accuracy)
j += 8
row += 1
save_path = os.getcwd() + "/Files/Combined/" + th + "/" + t
workbook_eval.save(save_path +
"/Evaluation_Parameters".encode('utf-8') +
".xls")
print("--- %s Minutes ---" % str((time.time() - start_time) / 60))
for word in rumor.split():
word = str(stemmer.stem(word))
if word not in lookedInWordnet:
lookedInWordnet.add(word)
syns = wn.synsets(word)
for s in syns:
tokens = s.lemma_names()
for token in tokens:
token = str(stemmer.stem(token))
if token not in wordnetDict:
wordnetDict[token] = word
parentDict[word] = 0
word_dict = parentDict
if k == 2:
print "d"
edge_dict = ut.tuples(word_array)
else:
edge_dict, wordnetDictGraph = ut.construct_graph(
rumor, wordnetDictGraph)
rumor_edge_count[k] = edge_dict
rumor_word_count[k] = word_dict
k += 1
threshold = np.arange(0, 1, 0.01)
original_rumor_word_count = rumor_word_count
original_rumor_edge_count = rumor_edge_count
word_count_storage = {}
for val in threshold:
i = 0
rumor_edge_count = original_rumor_edge_count
def main():
# if __name__ == "__main__":
print "Started---------------------------"
start_time = time.time()
thresh = ["10", "20"] #["10","15", "20", "25"]
workbook = xlrd.open_workbook(SETTINGS.dataset_folder + "dataset.xlsx")
stop_words = open(SETTINGS.stop_words_file).read().splitlines()
tweets, num_rows, tweetLabel, followerDict = ut.tweet_from_source_with_Followers(
workbook, stop_words)
rumor_category = {"NA", "R1", "R3", "R4", "R5", "R7", "All"}
rumor_type = {'Description'}
for th in thresh:
for t in sorted(rumor_type):
scores = {}
rumor_list = ut.get_rumor_list(workbook, t, stop_words)
rumor_edge_count = {i: {} for i in range(rumor_list.__len__())}
k = 0
for rumor in rumor_list:
word_array = rumor.split()
edge_dict = ut.tuples(word_array)
rumor_edge_count[k] = edge_dict
k += 1
threshold = np.arange(0, 0.5, 0.01)
original_rumor_edge_count = rumor_edge_count
for val in threshold:
i = 0
print "Threshold Now:::::::::::::::::::::::::::::::::::: " + str(
val)
# rumor_edge_count = original_rumor_edge_count
rumor_edge_count = original_rumor_edge_count
labeled_tweets, rumor_edge_count = ut.label_data_graph(
tweets, rumor_edge_count, val, rumor_list.__len__(), th,
followerDict)
workbook_new = xlwt.Workbook()
sheet = workbook_new.add_sheet("confusion Matrix")
sheet.write(0, 0, "Tweets")
sheet.write(0, 1, "Category")
sheet.write(0, 2, "True Value")
tp = {}
tn = {}
fp = {}
fn = {}
excel = {}
for key in labeled_tweets:
cat = None
if key != 0 and key != 2 and key != 3 and key != 4 and key != 6:
cat = "NA"
elif key == 0:
cat = "R1"
elif key == 2:
cat = "R3"
elif key == 3:
cat = "R4"
elif key == 4:
cat = "R5"
elif key == 6:
cat == "R7"
value = labeled_tweets[key]
if cat == "R3":
print value.__len__()
for item in value:
if item in tweetLabel:
type = tweetLabel[item]
if type != "NA" and cat != "NA":
if "All" not in tp:
tp["All"] = 1
else:
tp["All"] = tp["All"] + 1
elif type == "NA" and cat != "NA":
if "All" not in fp:
fp["All"] = 1
else:
fp["All"] = fp["All"] + 1
elif type != "NA" and cat == "NA":
if "All" not in fn:
fn["All"] = 1
else:
fn["All"] = fn["All"] + 1
if type == cat:
if type not in tp:
tp[type] = 1
else:
tp[type] = tp[type] + 1
else:
if cat not in fp:
fp[cat] = 1
else:
fp[cat] = fp[cat] + 1
if type not in fn:
fn[type] = 1
else:
fn[type] = fn[type] + 1
excel[item] = cat
for key in rumor_category:
tn[key] = tweets.__len__() - tp.get(key, 0) - fp.get(
key, 0) - fn.get(key, 0)
tptnlist = []
tptnlist.append(tp)
tptnlist.append(fp)
tptnlist.append(fn)
tptnlist.append(tn)
scores[val] = tptnlist
sorted_scores = sorted(scores.items(), key=operator.itemgetter(0))
# cPickle.dump(word_count_storage, open("word_count_storage_Graph_"+str(t)+".p", "wb"))
# cPickle.dump(scores, open("scores_Graph_"+t+".p", "wb"))
ut.calculations(sorted_scores, rumor_category, t, th, "Graph")
workbook_eval = xlwt.Workbook()
sheet = workbook_eval.add_sheet("Evaluation")
sheet.write(0, 0, "Threshold")
j = 0
for key in rumor_category:
if key == "R7" or key == "R5" or key == "R4":
continue
sheet.write(0, j + 1, key + "_True Positive")
sheet.write(0, j + 2, key + "_False positive")
sheet.write(0, j + 3, key + "_True Negative")
sheet.write(0, j + 4, key + "_False Negative")
sheet.write(0, j + 5, key + "_Precision")
sheet.write(0, j + 6, key + "_Recall")
sheet.write(0, j + 7, key + "_F1-Measure")
sheet.write(0, j + 8, key + "_Accuracy")
j += 8
row = 1
for score, tptnlist in sorted_scores:
sheet.write(row, 0, score)
# tptnlist = sorted_scores[val];
tp1 = tptnlist[0]
fp1 = tptnlist[1]
fn1 = tptnlist[2]
tn1 = tptnlist[3]
j = 0
for key in rumor_category:
if key == "R7" or key == "R5" or key == "R4":
continue
p_1 = tp1.get(key, 0) + fp1.get(key, 0)
precision = 0
if p_1 != 0:
precision = float(tp1.get(key, 0)) / (p_1)
recall = float(tp1.get(
key, 0)) / (tp1.get(key, 0) + fn1.get(key, 0))
sheet.write(row, j + 1, tp1.get(key, 0))
sheet.write(row, j + 2, fp1.get(key, 0))
sheet.write(row, j + 3, tn1.get(key, 0))
sheet.write(row, j + 4, fn1.get(key, 0))
sheet.write(row, j + 5, precision)
sheet.write(row, j + 6, recall)
if precision + recall == 0:
sheet.write(row, j + 7, 0)
else:
sheet.write(
row, j + 7,
float(2 * precision * recall /
(precision + recall)))
accuracy = float(
(tp1.get(key, 0) + tn1.get(key, 0)) / float(
tp1.get(key, 0) + tn1.get(key, 0) +
fp1.get(key, 0) + fn1.get(key, 0)))
sheet.write(row, j + 8, accuracy)
j += 8
row += 1
save_path = os.getcwd() + "/Files/Graph/" + th + "/" + t
workbook_eval.save(save_path +
"/Evaluation_Parameters".encode('utf-8') +
".xls")