def calculate_norm_indegree_entropy(user_graph):
out_degree_vector,\
out_weighted_degree_vector, \
in_degree_vector, \
in_weighted_degree_vector, \
total_degree_vector,\
total_weighted_degree_vector = get_degree_directed(user_graph)
number_of_comments = int(np.sum(in_weighted_degree_vector))
if number_of_comments < 2:
norm_indegree_entropy = 1.0
return norm_indegree_entropy
# if within_discussion_anonymous_coward is not None:
# number_of_users = len(set_of_users) - 1
# else:
# number_of_users = len(set_of_users)
# Calculate the user-to-comment distribution entropy.
replied_to_distribution = in_weighted_degree_vector / number_of_comments
replied_to_distribution = replied_to_distribution[
replied_to_distribution > 0]
number_of_users = replied_to_distribution.size
replied_to_entropy = np.abs(-np.sum(
np.multiply(replied_to_distribution, np.log(replied_to_distribution))))
# Calculate the maximum possible user-to-comment distribution entropy given the number of comments.
uniform_replied_to_count = np.zeros(number_of_users, dtype=np.float64)
uniform_replied_to_count += (number_of_comments // number_of_users)
uniform_replied_to_count[:(number_of_comments % number_of_users)] += 1
uniform_replied_to_distribution = uniform_replied_to_count / number_of_comments
uniform_replied_to_distribution = uniform_replied_to_distribution[
uniform_replied_to_distribution > 0]
max_replied_to_entropy = np.abs(-np.sum(
np.multiply(uniform_replied_to_distribution,
np.log(uniform_replied_to_distribution))))
# Calculate normalized user-to-comment entropy.
if max_replied_to_entropy == 0.0:
norm_indegree_entropy = 1.0
else:
norm_indegree_entropy = np.abs(replied_to_entropy /
max_replied_to_entropy)
return norm_indegree_entropy
def calculate_outdegree_entropy(user_graph):
out_degree_vector,\
out_weighted_degree_vector, \
in_degree_vector, \
in_weighted_degree_vector, \
total_degree_vector,\
total_weighted_degree_vector = get_degree_directed(user_graph)
number_of_comments = np.sum(out_weighted_degree_vector)
if number_of_comments < 2:
outdegree_entropy = 0
return outdegree_entropy
comment_distribution = out_weighted_degree_vector / number_of_comments
comment_distribution = comment_distribution[comment_distribution > 0]
outdegree_entropy = np.abs(-np.sum(
np.multiply(comment_distribution, np.log(comment_distribution))))
return outdegree_entropy
def calculate_indegree_entropy(user_graph):
out_degree_vector,\
out_weighted_degree_vector, \
in_degree_vector, \
in_weighted_degree_vector, \
total_degree_vector,\
total_weighted_degree_vector = get_degree_directed(user_graph)
number_of_comments = np.sum(in_weighted_degree_vector)
if number_of_comments < 2:
indegree_entropy = 0
return indegree_entropy
replied_to_distribution = in_weighted_degree_vector / number_of_comments
replied_to_distribution = replied_to_distribution[
replied_to_distribution > 0]
indegree_entropy = np.abs(-np.sum(
np.multiply(replied_to_distribution, np.log(replied_to_distribution))))
return indegree_entropy
def calculate_user_graph_hirsch(user_graph):
out_degree_vector,\
out_weighted_degree_vector, \
in_degree_vector, \
in_weighted_degree_vector, \
total_degree_vector,\
total_weighted_degree_vector = get_degree_directed(user_graph)
sorted_indices = np.argsort(out_weighted_degree_vector)
# This is the previous hirsch index value
minimum_hirsch_value = 1.0
user_graph_hirsch = 1.0
maximum_hirsch_value = out_weighted_degree_vector.size
if maximum_hirsch_value > minimum_hirsch_value:
comment_count = out_weighted_degree_vector[
sorted_indices[-maximum_hirsch_value]]
if comment_count >= maximum_hirsch_value:
user_graph_hirsch = maximum_hirsch_value
return user_graph_hirsch