时间序列神器Prophet: 如何实现突变点预测

import numpy as np
import pandas as pd
import os
import datetime 
import time
import re
np.random.seed(42)   

import plotly_express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import matplotlib.pyplot as plt
%matplotlib inline
# 设置支持中文字体
plt.rcParams['font.sans-serif'] = ['SimHei'] 
# 设置图像标题字体
plt.rcParams['axes.unicode_minus'] = False
import seaborn as sns
sns.set_theme(style="darkgrid")

import statsmodels.api as sm
import statsmodels.formula.api as smf
from statsmodels.tsa.stattools import adfuller
from sklearn.metrics import r2_score,mean_squared_error,mean_squared_log_error

from prophet import Prophet 

import warnings
warnings.filterwarnings('ignore')

df = pd.read_csv("https://raw.githubusercontent.com/facebook/prophet/main/examples/example_wp_log_peyton_manning.csv")
# df.to_csv("data.csv",index=False)  # 保存到本地
df

m = Prophet()
m.fit(df)
09:18:15 - cmdstanpy - INFO - Chain [1] start processing
09:18:16 - cmdstanpy - INFO - Chain [1] done processing

<prophet.forecaster.Prophet at 0x784fe27ee9e0>

future = m.make_future_dataframe(periods=365)  # 指定预测一年的数据
future

forecast = m.predict(future)
forecast.head()

forecast.columns  # 生成预测数据的全部字段信息

Index(['ds', 'trend', 'yhat_lower', 'yhat_upper', 'trend_lower', 'trend_upper',
       'additive_terms', 'additive_terms_lower', 'additive_terms_upper',
       'weekly', 'weekly_lower', 'weekly_upper', 'yearly', 'yearly_lower',
       'yearly_upper', 'multiplicative_terms', 'multiplicative_terms_lower',
       'multiplicative_terms_upper', 'yhat'],
      dtype='object')

from prophet.plot import add_changepoints_to_plot
fig = m.plot(forecast)
a = add_changepoints_to_plot(fig.gca(), m, forecast)

m = Prophet(changepoint_prior_scale=0.5)  # 默认值
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)

m = Prophet(changepoint_prior_scale=0.001)   
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)

m = Prophet(changepoint_prior_scale=1)   
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)

m = Prophet(changepoints=['2014-01-01'])  #  指定突变点的位置
forecast = m.fit(df).predict(future)
fig = m.plot(forecast)