try: # If model is already available in disk.
model,countVec,countMatrix,n_topic = train.recoverModel(H.modelname,H.datafolder)
beta_lda = model.components_/model.components_.sum(axis=1)[:, np.newaxis]
except:
n_topic = H.n_topic
#train_idx, valid_idx = train.splitData(H.df_red,0.1)
countVec, countMatrix = train.count(H.df_red,[1,2], 0.002,0.6)
model,beta_lda = train.fitLDA(countMatrix,n_topic=H.n_topic,random_state=6,alpha=None)
train.saveModel(H.modelname,H.datafolder,model,countVec,countMatrix,H.n_topic)
topic_dist, dominant_topic = train.transformer(countMatrix, model)
if __name__ == 'main':
# Recommender object.
R = Recommender(H.df_red,n_topic,topic_dist,H.links)
R.clearHistory() # R.history = np.array([],dtype='int64')
# search_result = R.searchPage('Orta Doğu Teknik Üniversitesi', True)
# search_result = R.searchPage('Makine Öğrenimi', True)
# search_result = R.searchPage('Suriye', True)
# search_result = R.searchPage('Python (programlama dili)', True)
# search_result = R.searchPage('General Dynamics F-16 Fighting Falcon', True)
# search_result = R.searchPage('Kadeş Antlaşması', True)
# search_result = R.searchPage('Cengiz Han', True)
search_result = R.searchPage('Wolfgang Amadeus Mozart', True)
R.addToHistory(search_result.index.to_numpy())
print('Lets get recommendations..')
rcm = R.recommendWrtHistory()
R.evaluateRecommendation(rcm.head(20))
