def core_PROPHET(df_master, dimension, region, changepoint_prior_scale,
periods):
df_master.rename(columns={
df_master.columns[0]: 'ds',
df_master.columns[1]: 'y'
},
inplace=True)
m = Prophet(changepoint_prior_scale=changepoint_prior_scale)
m.fit(df_master)
future = m.make_future_dataframe(periods=periods)
forecast = m.predict(future)
if region == 'all':
region = 'Italy'
fig = m.plot(forecast, xlabel='Date', ylabel=dimension + ' for ' + region)
add_changepoints_to_plot(fig.gca(), m, forecast)
m.plot_components(forecast)
plt.pause(10)
forecast.rename(columns={
'ds': 'data',
'yhat': 'predizione',
'yhat_lower': 'intervallo_inferiore',
'yhat_upper': 'intervallo_superiore'
},
inplace=True)
print(forecast[[
'data', 'predizione', 'intervallo_inferiore', 'intervallo_superiore'
]].tail(periods))
def get_figures(filepath):
df = pd.read_csv(filepath)
df.columns = ['ds', 'y']
m = Prophet().fit(df)
future = m.make_future_dataframe(periods=90)
forecast = m.predict(future)
forecast_fig = m.plot(forecast)
add_changepoints_to_plot(forecast_fig.gca(), m, forecast)
components_fig = m.plot_components(forecast)
return forecast_fig, components_fig
def process_job(conn, job):
assert job['type'].lower() in ['cases', 'deaths', 'tests']
print(f"{time.strftime('%H:%M:%S')} Starting job={job}")
data = query_data(conn, job)
df = prepare_data(job, data)
m = create_model(job)
m.fit(df)
# predict a third into the future of what we looked back
future_days = round(job['days_to_look_back'] / 3)
future = m.make_future_dataframe(periods=future_days)
future['cap'] = df['cap'][0]
forecast = m.predict(future)
# region debug
if os.getenv('DOCKER_ACTIVE') is None:
fig = m.plot(forecast)
add_changepoints_to_plot(fig.gca(), m, forecast)
fig.savefig(f"../job/prediction-{job['id']}.png")
# endregion
change_points = m.changepoints.dt.date.tolist()
store_prediction(conn, job, forecast, change_points)
# cross validate and create score
if job['with_score']:
# compute period to have 5-6 simulated forecasts
horizon = pd.Timedelta("14 days")
initial = horizon * 3
period = (df.iloc[-1]['ds'] - df.iloc[0]['ds'] - horizon - initial) / 5
df_cv = cross_validation(m,
initial=initial,
horizon=horizon,
period=period,
parallel='processes')
df_p = performance_metrics(df_cv)
# region debug
if os.getenv('DOCKER_ACTIVE') is None:
fig = plot_cross_validation_metric(df_cv, metric='mape')
fig.savefig(f"../job/score-{job['id']}.png")
# endregion
score = df_p.iloc[-1]['mape']
store_score(conn, job, score)
def prophet(data):
model = Prophet()
model.fit(data)
future = model.make_future_dataframe(periods=365)
forecast = model.predict(future)
forecast.tail()
fig1 = model.plot(forecast)
fig2 = model.plot_components(forecast)
add_changepoints_to_plot(fig1.gca(), model, forecast, threshold=0)
plt.show()
def test3():
"""
手动指定潜在突变点的位置，而非利用自动的突变点检测，可以使用changepoints 参数
m = Prophet(changepoints=['2014-01-01'])
"""
df = pd.read_csv('data/example_wp_log_peyton_manning.csv')
m = Prophet(changepoints=['2014-01-01'])
m.fit(df)
future = m.make_future_dataframe(periods=365)
forecast = m.predict(future)
fig = m.plot(forecast)
add_changepoints_to_plot(fig.gca(), m, forecast)
m.plot(forecast)
plt.show()
def get_data():
from dateutil.parser import parse
data_path = u'/Users/longguangbin/Work/Documents/SAAS/安踏线下/季节性/sample/duanku_sales.xls'
data = pd.read_excel(data_path)
data['week_sale'] = data['sku_num_sum'].rolling(window=7).sum()
data.loc[:5, ['week_sale']] = data[:6]['sku_num_sum'].cumsum()
data['week_day'] = data['commit_date'].apply(lambda x: parse(x).weekday())
data_week = data[data['week_day'] == 6]
data['commit_month'] = data['commit_date'].apply(lambda x: parse(x).strftime('%Y-%m'))
data_month = data.groupby(['commit_month']).agg({'sku_num_sum': 'sum'}).reset_index()
df = data.loc[:, ['commit_date', 'sku_num_sum']]
df.columns = ['ds', 'y']
df = data_month.loc[:, ['commit_month', 'sku_num_sum']]
df.columns = ['ds', 'y']
df = data_week.loc[:, ['commit_date', 'sku_num_sum']]
df.columns = ['ds', 'y']
# df = data_month['sku_num_sum'].values
# df = data_week['sku_num_sum'].values
m = Prophet(growth='logistic', changepoint_prior_scale=0.005, changepoint_range=0.9)
df['cap'] = np.max(df['y']) * 1.2
df['floor'] = 0
m.fit(df)
future = m.make_future_dataframe(periods=360)
future['cap'] = np.max(df['y']) * 1.2
future['floor'] = 0
forecast = m.predict(future)
fig = m.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), m, forecast)
m.plot_components(forecast)
def daily_cases_fb_forecast(df, d, imgdir, attribution, figsize=(14, 9)):
# fit model
log.info("# Fitting daily cases using Prophet")
cases = df.groupby(df.date)['cases'].sum().diff()
df = cases.to_frame().reset_index().fillna(0)
df.columns = ['ds', 'y']
m = Prophet(interval_width=0.95)
with SuppressStdout():
m.fit(df)
future = m.make_future_dataframe(periods=d)
pred = m.predict(future)
# plot result
plt.figure(figsize=figsize)
fig = m.plot(pred)
_ = add_changepoints_to_plot(fig.gca(), m, pred)
plt.xlabel('Date [YYYY-MM-DD]')
plt.ylabel('Total Cases')
plt.title(
'Knoxville Metro COVID19 Forecasted Daily New Cases -- Updated: {}'.
format(time_now()))
plt.annotate(attribution['text'], (0, 0), (0, -60),
xycoords='axes fraction',
textcoords='offset points',
fontsize=attribution['fsize'],
color=attribution['color'],
alpha=attribution['alpha'])
plt.tight_layout()
plt.savefig(os.path.join(imgdir, 'metro-cases-all-daily-forecasted.png'))
def forecasting(cps,period,dataframe):
from fbprophet.plot import add_changepoints_to_plot
changed=Prophet(changepoint_prior_scale=cps)
changed.fit(dataframe)
future=changed.make_future_dataframe(periods=period)
forecast=changed.predict(future)
fig=changed.plot(forecast)
p=add_changepoints_to_plot(fig.gca(),changed,forecast)
return changed
def test2():
"""
如果趋势的变化被过度拟合（即过于灵活）或者拟合不足（即灵活性不够），可以利用输入参数 changepoint_prior_scale
来调整稀疏先验的程度。默认下，这个参数被指定为 0.05 。
增加这个值，使趋势变化更加灵活。减少这个值会使灵活度降低
"""
df = pd.read_csv('data/example_wp_log_peyton_manning.csv')
# 拟合模型
m = Prophet(changepoint_prior_scale=0.5)
m.fit(df)
future = m.make_future_dataframe(periods=365)
forecast = m.predict(future)
fig = m.plot(forecast)
add_changepoints_to_plot(fig.gca(), m, forecast)
plt.show()
def evaluateModel(model, data, endDate=None, title=None):
model.fit(data["training"])
holdout_period = model.make_future_dataframe(periods=len(data["holdout"]),
include_history=False)
holdout_forecast = model.predict(holdout_period)
if endDate is None:
periods = len(data["holdout"])
else:
periods = (endDate - data["training"].ds.max()).days
all_period = model.make_future_dataframe(periods=periods,
include_history=True)
all_forecast = model.predict(all_period)
text = '{}Holdout MAPE: {:,.2f}%'.format(
"{} - ".format(title) if title is not None else "",
calcMAPE(data["holdout"].y, holdout_forecast.yhat))
plotHTML = plot(
{
"data": [
go.Scatter(x=data["all"]['ds'], y=data["all"]['y'], name='y'),
go.Scatter(
x=all_forecast['ds'], y=all_forecast['yhat'], name='yhat'),
go.Scatter(x=all_forecast['ds'],
y=all_forecast['yhat_upper'],
fill='tonexty',
mode='none',
name='upper'),
go.Scatter(x=all_forecast['ds'],
y=all_forecast['yhat_lower'],
fill='tonexty',
mode='none',
name='lower'),
],
"layout":
go.Layout(title=text)
},
output_type='div')
plotProphet = model.plot(all_forecast)
add_changepoints_to_plot(plotProphet.gca(), model, all_forecast)
plotComponents = model.plot_components(all_forecast)
return {
"plot": plotHTML,
"prophetplot": plotProphet,
"plotcomponents": plotComponents
}
def show_trend(ncov_df,
name=None,
variable="Confirmed",
n_changepoints=2,
future_perd=0,
plot_n=False,
**kwargs):
"""
Show trend of log10(@variable) using fbprophet package.
@ncov_df <pd.DataFrame>: the clean data
@variable <str>: variable name to analyse
- if Confirmed, use Infected + Recovered + Deaths
@n_changepoints <int>: max number of change points
@kwargs: keword arguments of select_area()
"""
# Data arrangement
df = select_area(ncov_df, **kwargs)
try:
df = df.loc[:, ["Date", variable]]
df.columns = ["ds", "y"]
# Log10(x)
warnings.resetwarnings()
with warnings.catch_warnings():
warnings.simplefilter("ignore")
df["y"] = np.log10(df["y"]).replace([np.inf, -np.inf], 0)
# fbprophet
model = Prophet(growth="linear",
daily_seasonality=False,
n_changepoints=n_changepoints)
model.fit(df)
future = model.make_future_dataframe(periods=future_perd)
forecast = model.predict(future)
# future1 = model.make_future_dataframe(periods=)
# forecast1 = model.predict(future1)
# Create figure
if (plot_n):
fig = model.plot(forecast)
_ = add_changepoints_to_plot(fig.gca(), model, forecast)
if name is None:
try:
name = f"{kwargs['places'][0][0]}: "
except Exception:
name = str()
else:
name = f"{name}: "
plt.title(f"{name}log10({variable}) over time and chainge points")
plt.ylabel(f"log10(the number of cases)")
plt.xlabel("")
return ({
'change_point': model.changepoints,
'predict': forecast[['ds', 'yhat']]
})
except Exception:
pass
return (None)
def plot_prophet_forecast(prophet_model, prophet_pred, save_dir=None):
'''
Plot Prophet model's forecast using the Prophet API, including changepoints
:param prophet_model: Fitted Prophet model
:param prophet_pred: A forecast from a Prophet model (result of a prophet.predict() call)
'''
fig = prophet_model.plot(prophet_pred)
ax = fig.gca()
add_changepoints_to_plot(ax, prophet_model, prophet_pred)
ax = fig.gca()
ax.set_xlabel('Date')
ax.set_ylabel('Consumption [m^3]')
fig.suptitle('Prophet Model Forecast', fontsize=15)
fig.tight_layout(pad=2, rect=(0, 0, 1, 0.95))
save_dir = cfg['PATHS'][
'FORECAST_VISUALIZATIONS'] if save_dir is None else save_dir
plt.savefig(save_dir + 'Prophet_API_forecast' +
datetime.datetime.now().strftime("%Y%m%d-%H%M%S") + '.png')
return
def fbp_plot(forecast, model, changepoint=False):
fig = model.plot(forecast,
xlabel='date',
ylabel='percent increase',
figsize=(8, 6))
if changepoint:
a = add_changepoints_to_plot(fig.gca(), model, forecast)
plt.show()
model.plot_components(forecast, figsize=(8, 6))
plt.show()
def test1():
"""
自动检测突变点
"""
df = pd.read_csv('data/example_wp_log_peyton_manning.csv')
m = Prophet(n_changepoints=5)
m.fit(df)
# 预测
future = m.make_future_dataframe(periods=365)
forecast = m.predict(future)
# 获取显著突变点的位置
# 默认情况下，只有在时间序列的前80%才会推断出突变点，以便有足够的长度来预测未来的趋势，并避免在时间序列的末尾出现过度拟合的波动。
# 这个默认值可以在很多情况下工作，但不是所有情况下都可以，可以使用changepoint_range参数进行更改。
# 例如，Python中的m = Prophet(changepoint_range=0.9)。这意味着将在时间序列的前90%处寻找潜在的变化点
fig = m.plot(forecast)
add_changepoints_to_plot(fig.gca(), m, forecast)
plt.show()
def plot_forecast(market, period, frequency):
model = fit_model_kwargs(market)
future = model.make_future_dataframe(periods=period, freq=frequency)
forecast = model.predict(future)
forecast[['yhat', 'yhat_lower']] = forecast[['yhat',
'yhat_lower']].clip(0, )
fig = model.plot_components(forecast)
fig = model.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), model, forecast)
def saveFig(name, yr):
fig = m.plot(forecast)
plt.xlabel("year")
plt.ylabel("CAD")
plt.title(name + " price forecast, weekly prices from " + str(yr) +
" to 2023")
plt.subplots_adjust(top=0.9)
plt.savefig(resultPath + name + str(term) + "1.png", dpi=800)
a = add_changepoints_to_plot(fig.gca(), m, forecast)
fig = m.plot_components(forecast)
plt.savefig(resultPath + name + str(term) + "2.png")
def prediction(self, period):
self.model.fit(self.prophet_df)
future = self.model.make_future_dataframe(periods=period)
# Eliminate weekend from future dataframe
future['day'] = future['ds'].dt.weekday
future = future[future['day'] <= 4]
forecast = self.model.predict(future)
fig = plot_plotly(self.model, forecast)
fig2 = self.model.plot_components(forecast)
fig3 = self.model.plot(forecast)
a = add_changepoints_to_plot(fig3.gca(), self.model, forecast)
return (fig, fig2, fig3)
def plot(self, changepoints=True):
"""
Plots actuals and forecasted values.
:param changepoints: Bool, whether or not to add changepoints and trendline to plot.
:return: Matplotlib Figure
"""
fig = self.m.plot(self.f.reset_index())
if changepoints:
add_changepoints_to_plot(fig.gca(), self.m, self.f.reset_index())
if self.acts is not None:
self.acts.plot(fig=fig,
ls='',
marker='D',
mec='black',
mfc='lightgray',
markersize=3)
return fig
def model_prophet1(data):
model = fbprophet.Prophet(changepoint_prior_scale=0.05,
daily_seasonality=True)
# 定义模型
model.fit(data) # 训练模型
forecast_df = model.make_future_dataframe(periods=365, freq='D') # 生成需预测序列
forecast = model.predict(forecast_df) # 模型预测
model.plot(forecast, xlabel='Date', ylabel='Close Price ¥') # 绘制预测图
plt.title('Close Price of 000001.SZ')
plt.show()
fig = model.plot(forecast) # 绘制预测图
a = add_changepoints_to_plot(fig.gca(), model, forecast) # 增加变化点
def corona_prdict():
df_korea = pd.read_csv('corona.csv')
df_prophet = df_korea[['date', 'patient']]
# df_prophet = df_korea.rename(columns={
# 'date': 'ds',
# 'confirmed': 'y'
# })
df_prophet.columns = ['ds', 'y']
# a = np.arange(0.8, 1.01, 0.1)
# b = np.arange(0.1, 0.3, 0.1)
#
# params_grid = {'changepoint_range': a,
# 'changepoint_prior_scale': b}
# grid = ParameterGrid(params_grid)
# for p in tqdm(grid):
# m = Prophet(**p,
# yearly_seasonality=False,
# weekly_seasonality=False,
# daily_seasonality=True,
# seasonality_mode='additive')
# m.fit(df_prophet)
# print('optimum parameter loaded !!')
m = Prophet(
changepoint_prior_scale=Main.scale, # 증가시 유연
changepoint_range=0.90, # 시계열의 첫 90%에 잠재적 변화점을 두다.
yearly_seasonality=False,
weekly_seasonality=False,
daily_seasonality=True,
seasonality_mode='additive')
m.fit(df_prophet)
future = m.make_future_dataframe(periods=7) # 7일만 예상하겠다.
forecast = m.predict(future)
import matplotlib.font_manager as fm
path = 'C:/Windows/Fonts/malgun.ttf'
font_name = fm.FontProperties(fname=path, size=50).get_name()
plt.rc('font', family=font_name)
fig = m.plot(forecast, figsize=(6, 4), xlabel='날 짜', ylabel='감염자수(예상)')
a = add_changepoints_to_plot(fig.gca(), m, forecast)
# fig2 = plot_plotly(m, forecast, xlabel='날 짜', ylabel='감염자수(예상)')
# py.iplot(fig2, filename='corona19.html')
date = time.strftime('%Y%m%d', time.localtime(time.time()))
# plotly.offline.plot(fig2, filename='corona_predict/corona19_{}.html'.format(date))
# plt.imshow(a)
plt.savefig(Main.PATH + "{}.png".format(date))
def get_custom_figures(filepath, txtFuture_periods, txtChangepoints,
txtChangepoint_range, txtChangepoint_scale,
ckbCountry_holidays, txtHolidays_scale,
ckbMonthly_seasonality, txtSeasonality_days,
txtFourier_monthly):
df = pd.read_csv(filepath)
df.columns = ['ds', 'y']
custom_figures = {
'n_changepoints': int(txtChangepoints),
'changepoint_range': float(txtChangepoint_range),
'changepoint_prior_scale': float(txtChangepoint_scale),
'holidays_prior_scale': float(txtHolidays_scale),
}
if ckbCountry_holidays:
ckbCountry = 'true'
else:
ckbCountry = 'false'
if ckbMonthly_seasonality:
ckbMonthly = 'true'
else:
ckbMonthly = 'false'
messsage = 'Future periods = ' + txtFuture_periods + ' || n_changepoints = ' + txtChangepoints + ' || changepoint_range = ' + txtChangepoint_range + ' || changepoint_range = ' + txtChangepoint_scale + ' || add_country_holidays(VN) = ' + ckbCountry + ' || holidays_prior_scale = ' + txtHolidays_scale + ' || add_seasonality(monnthly) = ' + ckbMonthly + ' || Seasonality days = ' + txtSeasonality_days + ' || Fourier monthly = ' + txtFourier_monthly
flash(messsage)
m = Prophet(**custom_figures)
if ckbCountry_holidays:
m.add_country_holidays(country_name='VN')
if ckbMonthly_seasonality:
m.add_seasonality(name='monthly',
period=float(txtSeasonality_days),
fourier_order=int(txtFourier_monthly))
m.fit(df)
future = m.make_future_dataframe(periods=int(txtFuture_periods))
forecast = m.predict(future)
forecast_fig = m.plot(forecast)
add_changepoints_to_plot(forecast_fig.gca(), m, forecast)
components_fig = m.plot_components(forecast)
return forecast_fig, components_fig
def prophet_analysis(df,split,freq,changepoints=3):
train = df.iloc[:split]
test = df.iloc[split:]
# m_eval = Prophet(growth='linear')
m_eval = Prophet(
growth='linear',
n_changepoints=changepoints,
changepoint_range=0.8,
yearly_seasonality=False,
weekly_seasonality=False,
daily_seasonality=False,
seasonality_mode='additive',
seasonality_prior_scale=20,
changepoint_prior_scale=.5,
mcmc_samples=0,
interval_width=0.8,
uncertainty_samples=500,
).add_seasonality(
name='monthly',
period=30.5,
fourier_order=5
).add_seasonality(
name='yearly',
period=365.25,
fourier_order=20
).add_seasonality(
name='quarterly',
period=365.25/4,
fourier_order=5,
prior_scale=15)
m_eval.fit(train)
eval_future=m_eval.make_future_dataframe(periods=test.shape[0],freq=freq)
eval_forecast=m_eval.predict(eval_future)
fig,axs=plt.subplots(1,1,figsize=(15,4))
ax1 = sns.lineplot(x='ds',y='yhat',data=eval_forecast,label='Predictions',legend='full')
ax1 = sns.lineplot(x='ds',y='y',data=train,label='Train True',legend='full',linestyle='-.')
ax1 = sns.lineplot(x='ds',y='y',data=test,label='Test True',legend='full')
ax =m_eval.plot(eval_forecast)
ax = add_changepoints_to_plot(fig.gca(),m_eval,eval_forecast)
predictions = eval_forecast.iloc[-test.shape[0]:]['yhat'] #grab predictions to compare with test set
print('MAPE = ' + str((abs(np.array(test.y)-predictions)/(np.array(test.y))).mean()))
print('RMSE = ' + str(rmse(predictions,test['y'])))
print('MEAN = ' + str(df.y.mean()))
return
def sav_figure(name, c_prophet, c_forecast, ca_cv=None):
fig = c_prophet.plot(c_forecast, xlabel='Date', ylabel='CPU Secs')
a = add_changepoints_to_plot(fig.gca(), c_prophet, c_forecast)
plt.title('CPU Seconds (hourly)')
plt.savefig(IMG_DIR + name + '_forecast.png')
#plt.show()
figC = c_prophet.plot_components(c_forecast)
plt.savefig(IMG_DIR + name + '_forecastComp.png')
#plt.show()
if ca_cv:
figV = plot_cross_validation_metric(ca_cv, metric='mape')
plt.title(
'CPU Seconds Forecast MAPE (mean absolute percent error)')
plt.savefig(IMG_DIR + name + '_forecastCV.png')
def plot_forecast(self):
"""Plot forecasted values"""
if self.residuals_forecast is not None:
fig, axes = super(ProphetForecaster, self)._plot_forecast(y=np.asarray(self._train_dt['y']),
yhat=np.asarray(self.fittedvalues['yhat']),
forecast=pd.Series(
np.asarray(self.forecast['yhat']),
index=self.forecast['ds']), _id='Prophet')
else:
fig_forecast = self._model.plot(self.forecast)
fig_components = self._model.plot_components(self.forecast)
if self._add_change_points:
a = add_changepoints_to_plot(fig_forecast.gca(), self._model, self.forecast)
plt.gcf().autofmt_xdate()
plt.grid(True)
plt.show()
def show_trend(ncov_df,
variable="Confirmed",
n_changepoints=2,
places=None,
excluded_places=None):
"""
Show trend of log10(@variable) using fbprophet package.
@ncov_df <pd.DataFrame>: the clean data
@variable <str>: variable name to analyse
- if Confirmed, use Infected + Recovered + Deaths
@n_changepoints <int>: max number of change points
@places <list[tuple(<str/None>, <str/None>)]: the list of places
- if the list is None, all data will be used
- (str, str): both of country and province are specified
- (str, None): only country is specified
- (None, str) or (None, None): Error
@excluded_places <list[tuple(<str/None>, <str/None>)]: the list of excluded places
- if the list is None, all data in the "places" will be used
- (str, str): both of country and province are specified
- (str, None): only country is specified
- (None, str) or (None, None): Error
"""
# Data arrangement
df = select_area(ncov_df, places=places, excluded_places=excluded_places)
if variable == "Confirmed":
df["Confirmed"] = df[["Infected", "Recovered", "Deaths"]].sum(axis=1)
df = df.loc[:, ["Date", variable]]
df.columns = ["ds", "y"]
# Log10(x)
warnings.resetwarnings()
with warnings.catch_warnings():
warnings.simplefilter("ignore")
df["y"] = np.log10(df["y"]).replace([np.inf, -np.inf], 0)
# fbprophet
model = Prophet(growth="linear",
daily_seasonality=False,
n_changepoints=n_changepoints)
model.fit(df)
future = model.make_future_dataframe(periods=0)
forecast = model.predict(future)
# Create figure
fig = model.plot(forecast)
_ = add_changepoints_to_plot(fig.gca(), model, forecast)
plt.title(f"log10({variable}) over time and chainge points")
plt.ylabel(f"log10(the number of cases)")
plt.xlabel("")
def plotter(train = None, title = "", ylabel = "", fn = "test", plot = True, add_chpts = False, add_cv = True):
#figure
fig = plt.figure(facecolor='w', figsize=(12, 7))
ax = fig.add_subplot(111)
plt.title(title, fontsize=18, y = 1.05)
### train result
train_fig = train.model.plot(train.forecast, ax = ax, xlabel = "Date", ylabel = ylabel)
if add_chpts:
cpts = add_changepoints_to_plot(train_fig.gca(), train.model, train.forecast)
ax = plt.gca(); ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
c_txt = "RMSE=%.1f MAPE=%.1f%%"%(train.param.rmse, train.cv_metrics.manual_mape)
#format date
myFmt = mdates.DateFormatter('%d-%m')
ax.xaxis.set_major_formatter(myFmt)
#ticks - locator puts ticks at regular intervals
xloc = plticker.MultipleLocator(base=3.0)
ax.xaxis.set_major_locator(xloc)
## Rotate date labels automatically
fig.autofmt_xdate()
xt.save(train_fig, xt.name(odir, fn+"_prediction" + ("_changepoints" if add_chpts else "") ), c_txt)
### components
train_comp_fig = train.model.plot_components(train.forecast)
ax = plt.gca(); ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
xt.save(train_comp_fig, xt.name(odir, fn+"_components"))
### MAPE
if add_cv:
train_cv_mape = train.cv_mape_fig()
ax = plt.gca()
ax.yaxis.set_major_formatter(FuncFormatter(xt.y_fmt))
xt.save(train_cv_mape, xt.name(odir, fn+"_cv_mape"))
plt.tight_layout()
if plot: plt.show()
plt.close('all')
def main():
df = pd.read_csv('./data/example_wp_log_peyton_manning.csv')
m = Prophet()
m.fit(df)
future = m.make_future_dataframe(periods=366)
forecast = m.predict(future)
fig = m.plot(forecast)
for cp in m.changepoints:
print(cp)
plt.axvline(cp, c='gray', ls='--', lw=2)
plt.show()
deltas = m.params['delta'].mean(0)
fig = plt.figure(facecolor='w', figsize=(10, 6))
ax = fig.add_subplot(111)
ax.bar(range(len(deltas)),
deltas,
facecolor='#0072B2',
edgecolor='#0072B2')
ax.grid(True, which='major', c='gray', ls='-', lw=1, alpha=0.2)
ax.set_ylabel('Rate change')
ax.set_xlabel('Potential changepoint')
fig.tight_layout()
fig.show()
fig = m.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), m, forecast)
fig.show()
m = Prophet(changepoint_prior_scale=0.5)
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)
fig.show()
m = Prophet(changepoint_prior_scale=0.001)
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)
fig.show()
m = Prophet(changepoints=['2014-01-01'])
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)
fig.show()
def test_change_point():
df = pd.read_csv(prophet_path + os.sep + 'example_wp_log_R.csv')
# 默认 first 80% of the time series in order to have plenty
# m = Prophet()
# m = Prophet(changepoint_range=0.9)
# flexibility 的调整？
# m = Prophet(changepoint_prior_scale=0.005)
# Specifying the locations of the changepoints
m = Prophet(changepoints=['2014-01-01'])
m.fit(df)
future = m.make_future_dataframe(periods=365)
# future.tail()
forecast = m.predict(future)
fig = m.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), m, forecast)
coin = 'bitcoin'
startDate = 20160201
endDate = 20200430
tables = pd.read_html('https://coinmarketcap.com/currencies/' + coin +
'/historical-data/?start=' + str(startDate) + '&end=' +
str(endDate))
#print(tables[0])
#reverse dataframe
table = tables[2].iloc[::-1]
#clean dataframe
df = table[['Date', 'Close**']].copy()
#normalize data
df = df.rename(index=str, columns={"Date": "ds", "Close**": "y"})
print(df)
#model
m = Prophet(seasonality_mode='multiplicative')
#m.add_seasonality('self_define_cycle',period=8,fourier_order=8,mode='additive')
m.fit(df)
future = m.make_future_dataframe(periods=365, freq='D')
fcst = m.predict(future)
fig = m.plot(fcst, xlabel='Date', ylabel='Closing value', uncertainty=True)
a = add_changepoints_to_plot(fig.gca(), m, fcst)
plt.title('Bitcoin forecast')
plt.show()
'''taking a 30 days future prediction with current model''' future_global=model.make_future_dataframe(periods=30,freq='D') # print(future_global.head()) # print(future_global.shape) '''predicting the future 30days data''' prediction = model.predict(future_global) # print(prediction) print(prediction[['ds','yhat','yhat_lower','yhat_upper']].tail()) '''plotting the predicted model graph(make as exponential)''' model.plot(prediction) model.plot_components(prediction) fig=model.plot(prediction) a=add_changepoints_to_plot(fig.gca(),model,prediction) '''checking the cross validation of the predicted model''' df_cv=cross_validation(model,horizon='30 days',period='15 days',initial='90 days') # print(df_cv.head()) # print(df_cv.shape) '''checking the performance metrics of the cross validated predicted model''' df_performance = performance_metrics(df_cv) # print(df_performance.head()) df_performance = plot_cross_validation_metric(df_cv,metric='rmse') df_performance = plot_cross_validation_metric(df_cv,metric='mse') df_performance = plot_cross_validation_metric(df_cv,metric='mae') df_performance = plot_cross_validation_metric(df_cv,metric='mape') df_performance = plot_cross_validation_metric(df_cv,metric='mdape')
