SARIMA的python实现，包括季节性分解和自动识别参数_sarima python 实现

import matplotlib.pyplot as plt
from statsmodels.tsa.statespace.sarimax import SARIMAX
import numpy as np
from pmdarima.arima import auto_arima
import statsmodels.api as sm
import pandas as pd
raw_data=pd.read_excel(r"C:\Users\附件4、土壤蒸发量2012—2022年.xls")
data=raw_data["土壤蒸发量(mm)"].values
index = pd.date_range(start='2012-01-01', periods=len(data), freq='M')
series=pd.Series(data,index=index)
 
diff1=series.diff(1)
diff2=series.diff(2)
print(u'原始序列的ADF检验结果为：',  ADF(series))
print(u'一阶差分序列的ADF检验结果为：',  ADF(diff1[1:]))
# 原始序列的ADF检验结果为： (-1.5045801917212067, 0.5312443790121629, 11, 111, {'1%': -3.490683082754047, '5%': -2.8879516565798817, '10%': -2.5808574442009578}, 642.5500224639288)
# 一阶差分序列的ADF检验结果为： (-10.061376456523105, 1.3307651257364255e-17, 10, 111, {'1%': -3.490683082754047, '5%': -2.8879516565798817, '10%': -2.5808574442009578}, 637.9083146867483)
#看输出的第二项，<=0.05则符合要求
 
 
#分解周期信息，得到周期
# decomposition = sm.tsa.seasonal_decompose(series, model='additive', extrapolate_trend='freq')
# fig = decomposition.plot()
# plt.show()
 
#自动求解参数
# model = auto_arima(series, seasonal=True, m=12, stepwise=True, trace=True,
#                    suppress_warnings=True, error_action="ignore", max_order=None)
# print(model.summary())
#SARIMAX(0, 0, 3)x(1, 0, [1], 12)
 
model = SARIMAX(series, order=(1,1,3), seasonal_order=(1, 0, 1, 12))
ar_result = model.fit()
# 打印模型的回归系数
# print("AR 模型的回归系数：")
# print(ar_result.params)
forecast_steps = 12  # 预测未来 3 个时间步
forecast = ar_result.get_prediction(start=len(data), end=len(data) + forecast_steps - 1)
forecast_mean = forecast.predicted_mean
forecast_ci = forecast.conf_int()
plt.figure(figsize=(10, 6))
plt.rcParams['font.sans-serif'] = ['Kaitt', 'SimHei']
plt.plot(series,label="观测值")
plt.plot(forecast_mean,linestyle='--',label="预测值")
plt.fill_between(forecast_ci.index, forecast_ci.iloc[:, 0], forecast_ci.iloc[:, 1], color='orange', alpha=0.2)
plt.title("时序预测图")
plt.legend()
plt.show()