def __train_topicmodel(self, num_topics=1):
if num_topics <= 1:
if self.num_topics == 1:
self.num_topics, optim_decay = self.__chooseK(limit=20,
start=2,
step=1)
else:
raise ValueError(
"Please enter num_topics '> 1' or put '= 1' if you want suggestions"
)
else:
# if self.num_topics == 1:
self.num_topics = num_topics
optim_decay = 0.7
if self.algo == 'gensim':
# Build LDA model
ldamodel = LdaMulticore(corpus=self.corpus,
id2word=self.dictionary,
num_topics=self.num_topics,
passes=20,
alpha='asymmetric',
eta='auto',
random_state=42,
iterations=1000,
per_word_topics=True,
eval_every=None)
elif self.algo == 'sklearn':
ldamodel = LatentDirichletAllocation(n_components=self.num_topics,
learning_method='batch',
max_iter=20,
learning_decay=optim_decay,
max_doc_update_iter=1000,
n_jobs=-1,
random_state=42)
lda_output = ldamodel.fit_transform(self.lda_dtm)
self.ldamodel = ldamodel
