def get_message_weight(self,body):
body_tdm = SimpleTermDocumentMatrix()
body_tdm.add_doc(body)
body_terms = body_tdm.get_terms(min_doc_freq=1)
tm = self.tdm.get_term_freq()
for body_term in body_terms:
tm_log10 = [math.log10(tm[k]) for k in tm if body_term == k]
mean = sum(tm_log10)/len(tm_log10) if len(tm_log10) > 0 else 1
return mean
def get_tdm(email_path) :
if os.path.isdir(email_path):
files = [ email_path + "/" + f for f in os.listdir(email_path) if os.path.isfile(os.path.join(email_path,f))]
else:
files = [email_path]
tdm = SimpleTermDocumentMatrix()
for f in files:
body = get_body(f)
tdm.add_doc(body)
return tdm
def train(self,files):
senders_map = {}
threads = {}
tdm = SimpleTermDocumentMatrix()
for f in files:
l = parse_email(f)
sender = l[0]
subject = l[1]
date_ = l[2]
body = l[3]
tdm.add_doc(body)
self.calc_sender_freq(sender,senders_map)
self.calc_weight_on_thread(subject,sender,date_,threads)
self.senders_freq = senders_map
self.threads_weight = threads
self.tdm = tdm