saymrwulf/prophet
1
0
Fork 0
mirror of https://github.com/saymrwulf/prophet.git synced 2026-07-23 19:32:25 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Refactoring mpl code to address #62, #63

Browse source
This commit is contained in:
Isaac Laughlin 2017-03-03 11:29:12 -08:00
parent 5cec062fce
commit e081db52e1
1 changed files with 91 additions and 73 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

164
python/fbprophet/forecaster.py
View file

@ -34,6 +34,8 @@ except ImportError:
# fb-block 2
class Prophet(object):
# Plotting default color for R/Python consistency
forecast_color = '#0072B2'
def __init__(
self,
growth='linear',
@ -616,16 +618,15 @@ class Prophet(object):
-------
a matplotlib figure.
"""
forecast_color = '#0072B2'
fig = plt.figure(facecolor='w', figsize=(10, 6))
ax = fig.add_subplot(111)
ax.plot(self.history['ds'].values, self.history['y'], 'k.')
ax.plot(fcst['ds'].values, fcst['yhat'], ls='-', c=forecast_color)
ax.plot(fcst['ds'].values, fcst['yhat'], ls='-', c=self.forecast_color)
if 'cap' in fcst:
ax.plot(fcst['ds'].values, fcst['cap'], ls='--', c='k')
if uncertainty:
ax.fill_between(fcst['ds'].values, fcst['yhat_lower'],
fcst['yhat_upper'], color=forecast_color, alpha=0.2)
fcst['yhat_upper'], color=self.forecast_color, alpha=0.2)
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.set_xlabel(xlabel)
ax.set_ylabel(ylabel)
@ -648,87 +649,104 @@ class Prophet(object):
a matplotlib figure.
"""
# Identify components to be plotted
plot_trend = True
plot_holidays = self.holidays is not None
plot_weekly = 'weekly' in fcst
plot_yearly = 'yearly' in fcst
components = [('plot_trend', True),
('plot_holidays', self.holidays is not None),
('plot_weekly', 'weekly' in fcst),
('plot_yearly', 'yearly' in fcst)]
components = [(plot, cond) for plot, cond in components if cond]
npanel = len(components)
fig, axes = plt.subplots(npanel, 1, facecolor='w', figsize=(9, 3 * npanel))
npanel = plot_trend + plot_holidays + plot_weekly + plot_yearly
forecast_color = '#0072B2'
fig = plt.figure(facecolor='w', figsize=(9, 3 * npanel))
panel_num = 1
ax = fig.add_subplot(npanel, 1, panel_num)
ax.plot(fcst['ds'].values, fcst['trend'], ls='-', c=forecast_color)
artists = []
for ax, plot in zip(axes, [getattr(self, plot) for plot, _ in components]):
artists += plot(fcst, ax=ax, uncertainty=uncertainty)
fig.tight_layout()
return artists
def plot_trend(self, fcst, ax=None, uncertainty=True, **plotargs):
artists = []
if not ax:
ax = fig.add_subplot(111)
artists += ax.plot(fcst['ds'].values, fcst['trend'], ls='-', c=self.forecast_color)
if 'cap' in fcst:
ax.plot(fcst['ds'].values, fcst['cap'], ls='--', c='k')
artists += ax.plot(fcst['ds'].values, fcst['cap'], ls='--', c='k')
if uncertainty:
ax.fill_between(
artists += [ax.fill_between(
fcst['ds'].values, fcst['trend_lower'], fcst['trend_upper'],
color=forecast_color, alpha=0.2)
color=self.forecast_color, alpha=0.2)]
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.xaxis.set_major_locator(MaxNLocator(nbins=7))
ax.set_xlabel('ds')
ax.set_ylabel('trend')
return artists
if plot_holidays:
panel_num += 1
def plot_holidays(self, fcst, ax=None, uncertainty=True):
artists = []
if not ax:
ax = fig.add_subplot(111)
holiday_comps = self.holidays['holiday'].unique()
y_holiday = fcst[holiday_comps].sum(1)
y_holiday_l = fcst[[h + '_lower' for h in holiday_comps]].sum(1)
y_holiday_u = fcst[[h + '_upper' for h in holiday_comps]].sum(1)
# NOTE the above CI calculation is incorrect if holidays overlap
# in time. Since it is just for the visualization we will not
# worry about it now.
artists += ax.plot(fcst['ds'].values, y_holiday, ls='-', c=self.forecast_color)
if uncertainty:
artists += [ax.fill_between(fcst['ds'].values, y_holiday_l, y_holiday_u,
color=self.forecast_color, alpha=0.2)]
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.xaxis.set_major_locator(MaxNLocator(nbins=7))
ax.set_xlabel('ds')
ax.set_ylabel('holidays')
return artists
def plot_weekly(self, fcst, ax=None, uncertainty=True):
artists = []
if not ax:
ax = fig.add_subplot(111)
df_s = fcst.copy()
df_s['dow'] = df_s['ds'].dt.weekday_name
df_s = df_s.groupby('dow').first()
days = pd.date_range(start='2017-01-01', periods=7).weekday_name
y_weekly = [df_s.loc[d]['weekly'] for d in days]
y_weekly_l = [df_s.loc[d]['weekly_lower'] for d in days]
y_weekly_u = [df_s.loc[d]['weekly_upper'] for d in days]
artists += ax.plot(range(len(days)), y_weekly, ls='-', c=self.forecast_color)
if uncertainty:
artists += [ax.fill_between(range(len(days)), y_weekly_l, y_weekly_u,
color=self.forecast_color, alpha=0.2)]
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.set_xticks(range(len(days)))
ax.set_xticklabels(days)
ax.set_xlabel('Day of week')
ax.set_ylabel('weekly')
return artists
def plot_yearly(self, fcst, ax=None, uncertainty=True):
artists = []
if not ax:
ax = fig.add_subplot(npanel, 1, panel_num)
holiday_comps = self.holidays['holiday'].unique()
y_holiday = fcst[holiday_comps].sum(1)
y_holiday_l = fcst[[h + '_lower' for h in holiday_comps]].sum(1)
y_holiday_u = fcst[[h + '_upper' for h in holiday_comps]].sum(1)
# NOTE the above CI calculation is incorrect if holidays overlap
# in time. Since it is just for the visualization we will not
# worry about it now.
ax.plot(fcst['ds'].values, y_holiday, ls='-', c=forecast_color)
if uncertainty:
ax.fill_between(fcst['ds'].values, y_holiday_l, y_holiday_u,
color=forecast_color, alpha=0.2)
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.xaxis.set_major_locator(MaxNLocator(nbins=7))
ax.set_xlabel('ds')
ax.set_ylabel('holidays')
df_s = fcst.copy()
df_s['doy'] = df_s['ds'].map(lambda x: x.strftime('2000-%m-%d'))
df_s = df_s.groupby('doy').first().sort_index()
artists += ax.plot(pd.to_datetime(df_s.index), df_s['yearly'], ls='-',
c=self.forecast_color)
if uncertainty:
artists += [ax.fill_between(
pd.to_datetime(df_s.index), df_s['yearly_lower'],
df_s['yearly_upper'], color=self.forecast_color, alpha=0.2)]
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
months = MonthLocator(range(1, 13), bymonthday=1, interval=2)
ax.xaxis.set_major_formatter(DateFormatter('%B %-d'))
ax.xaxis.set_major_locator(months)
ax.set_xlabel('Day of year')
ax.set_ylabel('yearly')
return artists
if plot_weekly:
panel_num += 1
ax = fig.add_subplot(npanel, 1, panel_num)
df_s = fcst.copy()
df_s['dow'] = df_s['ds'].dt.weekday_name
df_s = df_s.groupby('dow').first()
days = pd.date_range(start='2017-01-01', periods=7).weekday_name
y_weekly = [df_s.loc[d]['weekly'] for d in days]
y_weekly_l = [df_s.loc[d]['weekly_lower'] for d in days]
y_weekly_u = [df_s.loc[d]['weekly_upper'] for d in days]
ax.plot(range(len(days)), y_weekly, ls='-', c=forecast_color)
if uncertainty:
ax.fill_between(range(len(days)), y_weekly_l, y_weekly_u,
color=forecast_color, alpha=0.2)
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.set_xticks(range(len(days)))
ax.set_xticklabels(days)
ax.set_xlabel('Day of week')
ax.set_ylabel('weekly')
if plot_yearly:
panel_num += 1
ax = fig.add_subplot(npanel, 1, panel_num)
df_s = fcst.copy()
df_s['doy'] = df_s['ds'].map(lambda x: x.strftime('2000-%m-%d'))
df_s = df_s.groupby('doy').first().sort_index()
ax.plot(pd.to_datetime(df_s.index), df_s['yearly'], ls='-',
c=forecast_color)
if uncertainty:
ax.fill_between(
pd.to_datetime(df_s.index), df_s['yearly_lower'],
df_s['yearly_upper'], color=forecast_color, alpha=0.2)
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
months = MonthLocator(range(1, 13), bymonthday=1, interval=2)
ax.xaxis.set_major_formatter(DateFormatter('%B %-d'))
ax.xaxis.set_major_locator(months)
ax.set_xlabel('Day of year')
ax.set_ylabel('yearly')
fig.tight_layout()
return fig
# fb-block 9