def total_rmse():
group_count = DataReader.nym_count()
item_count = R.shape[1]
total_rmse = 0
item_lam = lam.sum(axis=0)
highest_n = 500
large_items = np.argpartition(item_lam, -highest_n)[-highest_n:]
with msg('Splitting group ratings'):
group_ratings = []
for group in range(group_count):
group_ratings.append(R[P[group]])
with msg('Getting rmse(s)'):
count = 0
for nth_item, item in enumerate(large_items):
for group in range(group_count):
mean = Rtilde[group, item]
# if mean < 3.5 and mean > 2.5:
# if mean > 4:
if True:
count += 1
data = group_ratings[group][:, item].data
var = Rvar[group, item]
if var == 0: var = 0.01
total_rmse += get_rmse(data, mean, var)
if (nth_item) % 10 == 0:
mean_rmse = total_rmse / (count)
print(f'[{nth_item}, {count}] Mean RMSE: {mean_rmse}')
def plot_nym_stat(thresh=thresh_default, inv=False, savefig=False, outfile=outfile_default, begin=None, num=None, stat_option=stat_option_default):
stat_name = stat_options[stat_option]
if inv: stat_name = f'inverse {stat_name}'
fig, ax = plt.subplots()
ax.set(
# ylim=(0, None),
title=f'{stat_name} of each group by item number (thresh no. ratings >= {thresh})',
xlabel='item number',
ylabel=stat_name)
cm = plt.get_cmap('gist_rainbow')
colors = [cm(1.*i/Data.nym_count()) for i in range(Data.nym_count())]
begin = 0 if begin is None else begin
end = None if num is None else begin + num
nym_stats = Data.get_nym_stats()[:, begin : (None if num is None else begin+num),:]
for nym_n in range(Data.nym_count()):
nym_n_stats = nym_stats[nym_n]
with msg(f'plotting nym #{nym_n} {stat_name}'):
valids = (nym_n_stats[:,3] >= thresh)
print(f'{valids.sum()} of {len(valids)} valid (thresh = {thresh})')
x = nym_n_stats[:,0][valids]
if stat_option is 1:
y = nym_n_stats[:,1][valids]
elif stat_option is 2:
y = nym_n_stats[:,2][valids]
elif stat_option is 3:
y = np.sqrt(nym_n_stats[:,2][valids])
if inv: y[y > 0] = 1 / y[y > 0]
s = np.sqrt(nym_n_stats[:,3][valids])
ax.scatter(x, y, s=s, facecolors='none', edgecolors=colors[nym_n], label=f'group {nym_n}')
ax.legend()
if savefig:
with msg('Saving "{}" to "{}"'.format(ax.title.get_text(), outfile)):
ax.get_figure().savefig(outfile, dpi=150)
plt.clf()
else:
plt.show()
import argparse
import os
from plot_item_dist import barplot_rating_dist
from datareader import DataReader as Data
parser = argparse.ArgumentParser(description="Plot distributions of each group for an item.")
parser.add_argument('item', help='item number to plot')
args = parser.parse_args()
item = args.item
fig_dir = f'figures/item{item}/'
if not os.path.exists(fig_dir):
os.makedirs(fig_dir)
for nym_n in range(Data.nym_count()):
barplot_rating_dist(item, group=nym_n, savefig=f'{fig_dir}item{item}_group{nym_n}_dist.png')
