Есть ли способ включить «накопительную» опцию, чтобы значение прогнозов никогда не уменьшалось?

df.set_index('Creation_date', inplace=True)
pd.to_datetime(df.index)
### Create one issue per row
df["Total_Issues"] = range(len(df))
df["Total_Issues"] = df["Total_Issues"] + 1
### Drops duplicate date values so we can create a time series
df = df.loc[~df.index.duplicated(keep='last'), :]
print("Removed duplicated dates")
print(df)
## Only Select the columns we need
df = df.loc[:,["Total_Issues"]]
## Fill empty dates so we get a full time series
df = df.resample('D').bfill()
### Predict Future ###
horizon = 180
forecast_df = df['Total_Issues']
last_date = forecast_df.index.max()
forecaster = Prophet()
forecaster.fit(forecast_df)
fh = ForecastingHorizon(
pd.date_range(str(last_date),
freq="D", periods = horizon),
is_relative=False
)
## Create prediction
y_pred = forecaster.predict(fh)
## Create confidence interval
ci = forecaster.predict_interval(fh, coverage=0.9)
## Plot
plt.show()
plt.plot(
forecast_df.tail(horizon*3),
label="Actual",
color="black"
)
plt.gca().fill_between(
ci.index, (ci.iloc[:,0]),
(ci.iloc[:,1]),
alpha=0.1,
)
plt.plot(y_pred, label="Predicted")
plt.ylim(bottom=0)
plt.legend()
plt.show()
Подробнее здесь: https://stackoverflow.com/questions/798 ... -in-python
Мобильная версия