# print number of clusters
print "Number of clusters:", len(list_of_clusters)
# map each test session to its keyword
list_test_keywords = map(lambda x : Keyword(x.keyword_id), list_of_test_sessions)
# compute predictions for CTR
list_CTR_predictions = []
count = 0
for kw in list_test_keywords:
print kw
for cluster in list_of_clusters:
if cluster.getKeyword(kw) != None:
print "FOUND"
list_CTR_predictions.append(cluster.computeCTR())
count += 1
break
print "found_ratio", count/float(len(list_test_keywords))
# Compute Mean Squared Error for predictions
m_s_e = evaluator(list_CTR_predictions)
print m_s_e
# remove the cluster we merged in
cluster_list.remove(secondCluster)
else:
continue
# Print out number of clusters:
print "Number of Clusters:", len(cluster_list)
# Compute predictions for keywords in test set
list_test_keywords = map(lambda x : Keyword(x.keyword_id), list_of_test_sessions)
list_CTR_predictions = []
list_CTR_weighted_predictions = []
for kw in list_test_keywords:
node = g_new.findNode(kw)
corresponding_cluster = node_cluster_dict[node]
list_CTR_predictions.append(corresponding_cluster.computeCTR())
list_CTR_weighted_predictions.append(corresponding_cluster.weightedCTR(node, g, g_new))
# Compute Mean Squared Error for predictions
m_s_e = evaluator(list_CTR_predictions)
print m_s_e
# Compute Mean Squared Error for predictions
m_s_e = evaluator(list_CTR_weighted_predictions)
print m_s_e