def plot_sim_costs(costs, savefig=None):
title = 'Program Outlay Validation Results'
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)
ys = costs['ri']
zs = costs['nj']
ws = costs['ca']
dct_color = dict(zip(sts, ['indianred', 'slategray', 'tan']))
ind = np.arange(len(ys))
width = 0.2
bar1 = ax.bar(ind-width, ys, width, align='center', capsize=5, color=dct_color['ri'], ecolor='grey')
bar2 = ax.bar(ind, zs, width, align='center', capsize=5, color=dct_color['nj'], ecolor='grey')
bar3 = ax.bar(ind+width, ws, width, align='center', capsize=5, color=dct_color['ca'], ecolor='grey')
ax.set_ylabel('Program Outlay')
ax.set_xticks(ind)
ax.set_xticklabels(('Logit', 'KNN', 'Naive Bayes', 'Random Forest', 'XGB', 'Ridge', 'SVC'))
ax.yaxis.grid(False)
ax.legend((bar1, bar2, bar3), ('Rhode Island','New Jersey', 'California' ))
ax.ticklabel_format(style='plain', axis='y')
# add horizontal bar for true numbers
dct_cost = dict(zip(sts, [166.7, 502.2, 5681.7]))
dct_offset = dict(zip(sts, [1.025, 1.025, 0.975]))
for st in sts:
y = dct_cost[st]
plt.axhline(y=y, color=dct_color[st], linestyle='--')
hline_offset = dct_offset[st]
hline_text = 'Actual Program Outlay, %s: %s million' % (st.upper(), y)
plt.text(2, y * hline_offset, hline_text, horizontalalignment='center', color='k')
format_chart(fig, ax, title, bg_color='white', fg_color='k')
if savefig is not None:
dir_out = savefig
# save
plt.savefig(dir_out + 'program_outlay.png', facecolor='white', edgecolor='grey') #
return None
def plot_ind_level(di, clf_class_names_plot, clf_profile, yvar, savefig=None):
# get clfs with cv results (ready for plot), dct_clf, and clf_class_names from clf_profile
clfs, clf_class_names_plot, dct_clf = clf_profile
# get clf class names, and corresponding labels for plot, must align for proper plotting
clf_class_names_plot = [type(z).__name__ for z in clfs]
clf_labels_plot = tuple([dct_clf[x] for x in clf_class_names_plot])
# make plot
title = 'Individual Level Validation Results - Performance Measures\nOutcome = %s' % yvar
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)
di = di[clf_class_names_plot] # order to ensure
ys = [x[yvar] for x in di.loc['precision'].values]
zs = [x[yvar] for x in di.loc['recall'].values]
ws = [x[yvar] for x in di.loc['f1'].values]
ind = np.arange(len(ys))
width = 0.2
bar1 = ax.bar(ind - width, ys, width, align='center', capsize=5, color='indianred', ecolor='grey')
bar2 = ax.bar(ind, zs, width, align='center', capsize=5, color='tan', ecolor='grey')
bar3 = ax.bar(ind + width, ws, width, align='center', capsize=5, color='slategray', ecolor='grey')
ax.set_ylabel('Performance Measure')
ax.set_xticks(ind)
ax.set_xticklabels(clf_labels_plot)
ax.yaxis.grid(False)
ax.legend((bar1, bar2, bar3), ('Precision', 'Recall', 'F1'))
ax.ticklabel_format(style='plain', axis='y')
# ax.get_yaxis().set_major_formatter(matplotlib.ticker.FuncFormatter(lambda x, p: format(int(x), ',')))
format_chart(fig, ax, title, bg_color='white', fg_color='k')
if savefig is not None:
dir_out, suffix = savefig
# save
plt.savefig(dir_out + 'ind_level%s.png' % suffix, facecolor='white', edgecolor='grey') #
return None
def plot_pop_level_worker_counts(dp, clf_class_names_plot, clf_profile, savefig=None):
'''
:param: clf_class_names: col of dp except 'true', numbers will follow this order in plot
'''
# get clfs with cv results (ready for plot), dct_clf, and clf_class_names from clf_profile
clfs, clf_class_names_plot, dct_clf = clf_profile
# get clf class names, and corresponding labels for plot, must align for proper plotting
clf_class_names_plot = [type(z).__name__ for z in clfs]
clf_labels_plot = tuple([dct_clf[x] for x in clf_class_names_plot])
# make plot
title = 'Population Level Validation Results - Worker Counts'
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)
dp = dp[clf_class_names_plot + ['true']]
ys = dp.drop(columns=['true']).loc['n_takers'].values
zs = dp.drop(columns=['true']).loc['n_needers'].values
ind = np.arange(len(ys))
width = 0.2
bar1 = ax.bar(ind - width / 2, ys, width, align='center', capsize=5, color='indianred', ecolor='grey')
bar2 = ax.bar(ind + width / 2, zs, width, align='center', capsize=5, color='tan', ecolor='grey')
ax.set_ylabel('Millions of workers')
ax.set_xticks(ind)
ax.set_xticklabels(clf_labels_plot)
ax.yaxis.grid(False)
ax.legend((bar1, bar2), ('Leave Takers', 'Leave Needers'))
ax.ticklabel_format(style='plain', axis='y')
ax.get_yaxis().set_major_formatter(matplotlib.ticker.FuncFormatter(lambda x, p: format(int(x), ',')))
format_chart(fig, ax, title, bg_color='white', fg_color='k')
# add horizontal bar for true numbers
n_takers_true = dp['true']['n_takers']
plt.axhline(y=n_takers_true, color='indianred', linestyle='--')
hline_offset = 1.025
hline_text = 'Actual Number of Takers: %s million' % (round(n_takers_true, 1))
plt.text(2, n_takers_true * hline_offset, hline_text, horizontalalignment='center', color='k')
n_needers_true = dp['true']['n_needers']
plt.axhline(y=n_needers_true, color='tan', linestyle='--')
hline_offset = 1.025
hline_text = 'Actual Number of Needers: %s million' % (round(n_needers_true, 1))
plt.text(2, n_needers_true * hline_offset, hline_text, horizontalalignment='center', color='k')
if savefig is not None:
dir_out, suffix = savefig
# save
plt.savefig(dir_out + 'pop_level_workers%s.png' % suffix, facecolor='white', edgecolor='grey') #
return None
def create_chart(self, out, sim_num):
output_directory = self.output_directories[sim_num]
# Plot costs and ci
total_cost = round(list(out.loc[out['type'] == 'total', 'cost'])[0] / 10 ** 6, 1)
spread = round((list(out.loc[out['type'] == 'total', 'ci_upper'])[0] -
list(out.loc[out['type'] == 'total', 'ci_lower'])[0]) / 10 ** 6, 1)
title = 'State: %s. Total Benefits Cost = $%s million (\u00B1%s).' % (self.st.upper(), total_cost, spread)
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)
ind = np.arange(6)
ys = out[:-1]['cost'] / 10 ** 6
es = 0.5 * (out[:-1]['ci_upper'] - out[:-1]['ci_lower']) / 10 ** 6
width = 0.5
ax.bar(ind, ys, width, yerr=es, align='center', capsize=5, color='#1aff8c', ecolor='white')
ax.set_ylabel('$ millions')
ax.set_xticks(ind)
ax.set_xticklabels(('Own Health', 'Maternity', 'New Child', 'Ill Child', 'Ill Spouse', 'Ill Parent'))
ax.yaxis.grid(False)
format_chart(fig, ax, title)
plt.savefig('%s/total_cost_%s_%s_%s' % (output_directory, self.yr, self.st, self.out_id),
facecolor='#333333', edgecolor='white')
self.figure = fig
return fig
'New Born Child',
'Ill Child',
'Ill Spouse',
'Ill Parent',
))
ax.yaxis.grid(False)
ax.legend((
bar1,
bar2,
), (
'Simulated',
'Actual',
))
ax.ticklabel_format(style='plain', axis='y')
#ax.get_yaxis().set_major_formatter(matplotlib.ticker.FuncFormatter(lambda x, p: format(int(x), ',')))
format_chart(fig, ax, title, bg_color='white', fg_color='k')
# save
plt.savefig(fp_out + 'RI_validation_n_cases.png',
facecolor='white',
edgecolor='grey') #
# plot for outlay
fp_out = './output/'
title = '' # 'Comparison of Simulation Results vs Program Data, Program Outlay in RI'
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)
ys = list(sim['outlay'].values())[-3:]
zs = list(actual['outlay'].values())[-3:]
ind = np.arange(len(ys))
width = 0.2
bar1 = ax.bar(ind - width / 2,
ys,
ind = np.arange(len(types))
ys = dct_bene.values()
width = 0.5
ax.bar(ind,
ys,
width,
align='center',
capsize=5,
color='#1aff8c',
ecolor='white')
ax.set_ylabel('Number of workers')
ax.set_xticks(ind)
ax.set_xticklabels(('Own Health', 'Maternity', 'New Child', 'Ill Child',
'Ill Spouse', 'Ill Parent'))
ax.yaxis.grid(False)
format_chart(fig, ax, title)
plt.savefig('./draft/demo_external_20191213/MD_RI_rf_low_inc_bene_counts.png',
facecolor='#333333',
edgecolor='white') #
## How much benefit would go to low-income worker families?
# set annual benefit = 0 for missing amount
for t in types:
acs.loc[acs['annual_benefit_%s' % t].isna(), 'annual_benefit_%s' % t] = 0
acs['annual_benefit_all'] = [
x.sum() for x in acs[['annual_benefit_%s' % t for t in types]].values
]
## Plot
# Number of benes, by benefit amount
title = 'Number of low-income worker recipients, by benefit levels'
fig, ax = plt.subplots(figsize=(8, 6), dpi=100)