def make_scatter(congress_num):
fig = plt.figure(1)
arr = hs.make_similarity_array(congress_num)
cls, means, steps = mlpy.kmeans(arr, k=2, plus=True)
members = hs.members(congress_num)
extreme_index1 = members.index(most_least.most_extreme(congress_num, 10)[0])
extreme_index2 = list(arr[extreme_index1]).index(min(arr[extreme_index1]))
if members[extreme_index1].split(' ')[1] == 'R':
#change color so Democrats are blue and Republicans are red
for i in range(len(cls)):
if cls[i] == 0:
cls[i] = 1
else:
cls[i] = 0
plot = plt.scatter(arr[:, extreme_index1], arr[:, extreme_index2], c=cls, alpha=0.75)
plt.xlabel("Conservatism (Cosine similarity to most conservative member, "
+ members[extreme_index1].split(' ')[0] + ")")
plt.ylabel("Liberalism (Cosine similarity to most liberal member, "
+ members[extreme_index2].split(' ')[0] + ")")
else:
plot = plt.scatter(arr[:, extreme_index2], arr[:, extreme_index1], c=cls, alpha=0.75)
plt.ylabel("Liberalism (Cosine similarity to most liberal member, "
+ members[extreme_index1].split(' ')[0] + ")")
plt.xlabel("Conservatism (Cosine similarity to most conservative member, "
+ members[extreme_index2].split(' ')[0] + ")")
return
def most_unique(congress_num, knn):
arr = hs.make_similarity_array(congress_num)
congresspeople = hs.members(congress_num)
#takes k-nearest members in cos similarity
sorted_similarity = [sorted(arr[i])[-knn:] for i in range(len(arr[0]))]
sums = [sum(sorted_similarity[i]) for i in range(len(sorted_similarity))]
sorted_unique = sorted(range(len(sums)), key=lambda k: sums[k])
unique_rank = [(congresspeople[index]) for index in sorted_unique]
return unique_rank
def most_extreme(congress_num, knn):
arr = hs.make_similarity_array(congress_num)
congresspeople = hs.members(congress_num)
#takes k-furthest members in cos similarity
sorted_similarity = [sorted(arr[i])[:knn] for i in range(len(arr[0]))]
#sums those similarities
sums = [sum(sorted_similarity[i]) for i in range(len(sorted_similarity))]
#sorts members from lowest to highest
sorted_extreme = sorted(range(len(sums)), key=lambda k: sums[k])
extreme_rank = [(congresspeople[index]) for index in sorted_extreme]
return extreme_rank
def nadler_index(congress_num, knn):
arr = hs.make_similarity_array(congress_num)
unique_rank = most_unique(congress_num, knn)
nadler_array = [arr[index][303] for (index, congressperson) in unique_rank]
return nadler_array