ARIMA预测每日进货价变化趋势

classname = '食用菌'
goal='批发价格(元/千克)'

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
plt.style.use('fivethirtyeight')
from matplotlib.pylab import rcParams
rcParams['figure.figsize'] = 28, 18
import statsmodels.api as sm
from statsmodels.tsa.stattools import adfuller
from statsmodels.tsa.seasonal import seasonal_decompose
import itertools
import warnings
warnings.filterwarnings("ignore")
import matplotlib.font_manager as fm

from pylab import mpl
mpl.rcParams['font.sans-serif'] = ['Microsoft YaHei'] # 指定默认字体：解决plot不能显示中文问题
mpl.rcParams['axes.unicode_minus'] = False # 解决保存图像是负号'-'显示为方块的问题

dateparse = lambda x: pd.to_datetime(x, format='%Y%m', errors = 'coerce')
df = pd.read_excel("品类数据总表（不含公式）.xlsx")

ts = df[pd.Series(pd.to_datetime(df.index, errors='coerce')).notnull().values]
ts.info()

print(ts)

ts.set_index('销售日期', inplace=True)
ts.sort_index(inplace=True)

ts.sort_index(inplace=True)

ts = ts[(ts[goal] > ts[goal].quantile(0.1)) & (ts[goal] < ts[goal].quantile(0.9))]

Energy_sources = ts.groupby('分类名称') #对排放数据进行分组
rcParams['figure.figsize'] = 28, 18

# Energy_sources = ts.groupby('分类名称') #对排放数据进行分组
# rcParams['figure.figsize'] = 28, 18
# fig, ax = plt.subplots()
# for desc, group in Energy_sources:
#     group.plot(y = goal, label = desc,ax = ax,fontsize = 25)
#     ax.set_xlabel(u'时间（月）')
#     ax.set_ylabel(goal)
#     ax.title.set_size(28)
#     ax.xaxis.label.set_size(25)
#     ax.yaxis.label.set_size(25)
#     ax.legend(loc = 'best', fontsize = 18)

# fig, axes = plt.subplots(3,2, figsize = (30, 20))
# for (desc, group), ax in zip(Energy_sources, axes.flatten()):
#     group.plot(y = goal,ax = ax, title = desc, fontsize = 25)
#     ax.set_xlabel('时间')
#     ax.set_ylabel(u'分类名称')
#     ax.xaxis.label.set_size(23)
#     ax.yaxis.label.set_size(23)

ts_grouped = ts.groupby('分类名称')
ts_classname = ts_grouped.get_group(classname)
print(ts_classname)

import pandas as pd

mte = ts_classname[goal]
mte = mte.fillna(0)
print(mte)

# 对mte作n阶差分，并作图
n = 1  # 阶数
diff_mte = mte.diff(n)
diff_mte.plot(figsize=(12, 6))
plt.xlabel('时间')
plt.ylabel('差分值')
plt.title('mte的{}阶差分'.format(n))
plt.show()

# adfuller单位根检验数据平稳性
from statsmodels.tsa.stattools import adfuller
print(adfuller(mte))  # 原始数据
diff_mte = mte.diff(1)
diff_mte = diff_mte.fillna(0)
print(adfuller(diff_mte))  # 一阶差分



diff_mte = mte.diff(2)
diff_mte = diff_mte.fillna(0)
print(adfuller(diff_mte))  # 二阶差分

# # 画pacf图和acf图
# from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
# pacf = plot_pacf(diff_mte, lags=40)
# plt.title('PACF')
# pacf.show()
# acf = plot_acf(diff_mte, lags=40)
# plt.title('ACF')
# acf.show()

# # 这里 max_ 数字越大运行时间越长，这里取8运行时间为 13m 13s
# train_results = sm.tsa.arma_order_select_ic(diff_mte, ic=['aic', 'bic'], trend='c', max_ar=8, max_ma=8)
# print('AIC', train_results.aic_min_order)
# print('BIC', train_results.bic_min_order)

from statsmodels.tsa.arima.model import ARIMA
# ARIMA(data, order=(p, d, q))
model = ARIMA(diff_mte, order=(6,2,4))   #经过模型验证，取q=6,p=2为arima的取值
result = model.fit()
result.summary()

import pandas as pd
import statsmodels.api as sm

# 提取关键信息
summary = result.summary()  # 获取结果的概要信息
params = result.params  # 获取模型参数
resid = result.resid  # 获取残差
fitted_values = result.fittedvalues  # 获取拟合值

# 创建一个 DataFrame
df = pd.DataFrame({
    'Summary': [summary],
    'Parameters': [params],
    'Residuals': [resid],
    'Fitted Values': [fitted_values]
})
import os
if not os.path.exists('arima result'):
    os.makedirs('arima result')
# 将数据存储到 Excel 文件
filename = 'arima result' + '/' + classname +'model'+ '.xlsx'
df.to_excel(filename, index=False)

print("ARIMA 结果已成功保存到 Excel 文件：", filename)

# # 获取残差
# resid = result.resid

# # 画qq图
# from statsmodels.graphics.api import qqplot
# qqplot(resid, line='q', fit=True)
# plt.show()

from scipy.stats import shapiro
shapiro(resid)

import statsmodels.api as sm
print(sm.stats.durbin_watson(resid.values))

# 预测出来的数据也为一阶差分
# predict(起始时间，终止时间)
predict = result.predict('2020-07-01','2023-06-30')
plt.figure(figsize=(12, 8))
plt.plot(diff_mte)
plt.plot(predict)

# 根据result，使用得到的arima模型预测未来7天的差分值，并且恢复原值
forecast_diff = result.forecast(steps=7)
print(forecast_diff)
forecast = mte.iloc[-1] + forecast_diff.cumsum()
print(forecast)

# Save the result to an xlsx file
import os
if not os.path.exists('arima result'):
    os.makedirs('arima result')
filename = 'arima result' + '/' + classname +'result'+ '.xlsx'
forecast.to_excel(filename, index=False)
print("ARIMA 结果已成功保存到 Excel 文件：", filename)