def classify_problem(self):
def __get_gp(x, item):
sd = item['sd']
avg = item['avg']
coe = 1 / (math.sqrt(2 * math.pi) * sd)
ind = (math.e)**(-((x-avg)*(x-avg)) / (2 * sd * sd))
return coe * ind
logging("Start classify problems...", 0)
for label in self.PROBLEM_VECTOR:
max_cla = ""
max_val = 0
for cla in self.CLASSIFY_AVG_SD:
if cla == u'others_sub':
continue
cla_p = 1.0
for i in range(0, self.VEC_CNT):
x = self.PROBLEM_VECTOR[label][i]
try:
p = __get_gp(x, self.CLASSIFY_AVG_SD[cla][i])
except Exception, ex:
logging("label: %s, cla: %s, dx: %s, error: %s" %(
label, cla, i), 2)
if p == 0: p = 0.00001
cla_p *= p
try:
cla_p *= self.CLASSIFY_P[cla]
except:
cla_p = 0
if cla_p > max_val:
max_val = cla_p
max_cla = cla
if label not in self.CLASSIFY_RES:
self.CLASSIFY_RES[label] = []
self.CLASSIFY_RES[label].append(max_cla)
def naive_bayes_classifier(repo, classify_list, classify_p, knowledge, vector_file):
logging("Problem Classify Program Start", 0)
problem_classify = ProblemClassify(repo, classify_list, classify_p, knowledge, vector_file)
problem_classify.read_problem_vector()
problem_classify.calculate_knowledge_avg_sd()
problem_classify.classify_problem()
logging("Problem Classify Program Finish", 0)
return problem_classify.CLASSIFY_RES
def get_user_aclist(self, step):
logging("Get all users in %s" % (self.REPO, ), 0)
users_list = []
with connection.cursor() as ues_con:
sql = "SELECT `user` FROM `ues_ojstatus` where `repo`=%s group by `user`"
ues_con.execute(sql, (self.REPO,))
users_tuple = ues_con.fetchall()
users_list = list(user[0] for user in users_tuple)
sql = "SET GLOBAL group_concat_max_len=%s"
ues_con.execute(sql, (step*10000,))
users_tuple = ues_con.fetchone()
logging("There are %d users in the %s DataBase" % (len(users_list), self.REPO), 0)
return users_list
def read_problem_vector(self):
logging("Start read the problem vector file...", 0)
with open(self.PROBLEM_VECTOR_FILE, 'r') as file:
self.PROBLEM_VECTOR = {}
pro_cnt, vec_cnt = map(int, file.readline().split(" "))
self.VEC_CNT = vec_cnt
for i in range(0, pro_cnt):
raw_data = file.readline().split(" ")
if(len(raw_data) == 1 and raw_data[0] == ""):
continue
self.PROBLEM_VECTOR[raw_data[0]] = map(float, raw_data[1:])
logging("Read the problem vector file finish", 0)
def calculate_knowledge_avg_sd(self):
logging('Start calculate knowledge avg sd...', 0)
for cla in self.KNOWLEDGE:
self.CLASSIFY_AVG_SD[cla] = []
for i in range(0, self.VEC_CNT):
cla_xi = []
for label in self.KNOWLEDGE[cla]:
try:
cla_xi.append(self.PROBLEM_VECTOR[label][i])
except Exception, ex:
logging("cla:%s, label:%s, error:%s" %(cla, label, ex), 2)
cla_xi = np.array(cla_xi)
data = {'avg': np.mean(cla_xi), 'sd': np.std(cla_xi)}
self.CLASSIFY_AVG_SD[cla].append(data)
def make_problem_vector(self):
logging("Start make problem vector...", 0)
sentences = word2vec.Text8Corpus(self.USER_ACLIST_FILE)
# sentences = word2vec.LineSentence(self.USER_ACLIST_FILE)
model = word2vec.Word2Vec(
sentences = sentences,
hs = 0,
sg = 0,
iter = 10,
size = 300,
window = 5,
sample = 1e-4,
)
model.wv.save_word2vec_format(fname=self.PROBLEM_VECTOR_FILE, binary=False)
logging("Make problem vector finish", 0)
class ProblemClassify(object):
def __init__(self, repo, classify_list, classify_p, knowledge, vector_file):
super
(ProblemClassify, self).__init__()
self.REPO = repo
self.VEC_CNT = 0
self.CLASSIFY_RES = {}
self.PROBLEM_VECTOR = {}
self.KNOWLEDGE = knowledge
self.CLASSIFY_P = classify_p
self.CLASSIFY_LIST = classify_list
self.PROBLEM_VECTOR_FILE = vector_file
# {
# 'class1': [{
# 'avg': xxx,
# 'sd'
# },
# {
# ...
# }
# ],
# 'class2': [{
# }
# ]
# }
self.CLASSIFY_AVG_SD = {}
def read_problem_vector(self):
logging("Start read the problem vector file...", 0)
with open(self.PROBLEM_VECTOR_FILE, 'r') as file:
self.PROBLEM_VECTOR = {}
pro_cnt, vec_cnt = map(int, file.readline().split(" "))
self.VEC_CNT = vec_cnt
for i in range(0, pro_cnt):
raw_data = file.readline().split(" ")
if(len(raw_data) == 1 and raw_data[0] == ""):
continue
self.PROBLEM_VECTOR[raw_data[0]] = map(float, raw_data[1:])
logging("Read the problem vector file finish", 0)
def calculate_knowledge_avg_sd(self):
logging('Start calculate knowledge avg sd...', 0)
for cla in self.KNOWLEDGE:
self.CLASSIFY_AVG_SD[cla] = []
for i in range(0, self.VEC_CNT):
cla_xi = []
for label in self.KNOWLEDGE[cla]:
try:
cla_xi.append(self.PROBLEM_VECTOR[label][i])
except Exception, ex:
logging("cla:%s, label:%s, error:%s" %(cla, label, ex), 2)
cla_xi = np.array(cla_xi)
data = {'avg': np.mean(cla_xi), 'sd': np.std(cla_xi)}
self.CLASSIFY_AVG_SD[cla].append(data)
logging('Calculate knowledge avg sd finish', 0)
def get_all_user_aclist(self):
logging("Start get all user aclist...", 0)
step = 1000
filename = os.path.join(TMP_CHDIR, '%s_label_list.txt'%(self.REPO, ))
user_list = self.get_user_aclist(step)
for i in range(0, len(user_list), step):
users = user_list[i: i+step]
users_label = self.fetch_user_lables(users)
self.write_user_aclist(users_label, self.USER_ACLIST_FILE)
logging('Users list %d--%d Done!'%(i, i+step), 0)
logging("Get all user aclist finish", 0)
def make_problems_vector(repo, aclist_file, vector_file):
logging("Make Problems Vectors Main Program Startup...", 0)
problem_vector = ProblemVector(repo, aclist_file, vector_file)
problem_vector.get_all_user_aclist()
problem_vector.make_problem_vector()
logging("Make Problems Vectors Finish", 0)
x = self.PROBLEM_VECTOR[label][i]
try:
p = __get_gp(x, self.CLASSIFY_AVG_SD[cla][i])
except Exception, ex:
logging("label: %s, cla: %s, dx: %s, error: %s" %(
label, cla, i), 2)
if p == 0: p = 0.00001
cla_p *= p
try:
cla_p *= self.CLASSIFY_P[cla]
except:
cla_p = 0
if cla_p > max_val:
max_val = cla_p
max_cla = cla
if label not in self.CLASSIFY_RES:
self.CLASSIFY_RES[label] = []
self.CLASSIFY_RES[label].append(max_cla)
# print self.CLASSIFY_RES
logging("Classify problems finish", 0)
def naive_bayes_classifier(repo, classify_list, classify_p, knowledge, vector_file):
logging("Problem Classify Program Start", 0)
problem_classify = ProblemClassify(repo, classify_list, classify_p, knowledge, vector_file)
problem_classify.read_problem_vector()
problem_classify.calculate_knowledge_avg_sd()
problem_classify.classify_problem()
logging("Problem Classify Program Finish", 0)
return problem_classify.CLASSIFY_RES
