使用Python的VAR模型
实际上这个代码的话也是有借鉴了其他文章上的啦。
但是我这边就着自己的项目修改了一下
想跟大家分享一下
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import warnings
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
import statsmodels.api as stm
import statsmodels.stats.diagnostic as dia
def GetClasses():
Data = pd.read_excel("data/info/USA.xlsx")
classNames = pd.value_counts(Data['Product Group']).index.array
return Data, classNames
def ReadTariffData(Data: pd.DataFrame, Name: str):
Data = Data[Data['Product Group'] == Name][:]
# 要加上inplace=True 才能在原数据上修改
Data.rename(index=Data["Indicator"], inplace=True)
DropItem = ['Partner Name', 'Reporter Name', 'Trade Flow',
'Product Group', 'Indicator',
'1993', '1994']
Data.drop(DropItem, axis=1, inplace=True)
Data = Data.stack().unstack(0)
Data = Data.dropna(axis=0, how='any')
DropItem = ['AHS AVE Tariff Lines Share (%)']
Data.drop(DropItem, axis=1, inplace=True)
return Data
def DataProcesser(Data: pd.DataFrame):
for name in Data.columns:
if name.find("%") < 0:
Data[name] = Data[name].apply(np.log)
return Data
def TimeSeriesPloter(Data: pd.DataFrame, name):
idx = 1
plt.figure("trend", figsize=(16, 16))
for item_idx, item in Data.iteritems():
# print(item_idx,'\n',item,'\n',type(item))
plt.subplot(2, 3, idx)
item.plot()
plt.title(item_idx)
idx += 1
plt.savefig('data/pics/' + '/timeseries' + name + '.png')
plt.show()
pass
def AcfPloter(Data: pd.DataFrame):
for item_idx, item in Data.iteritems():
# 绘制自相关图
plot_acf(item, auto_ylims=True).show()
plt.savefig('data/pics/' + '/AcfPlot_' + str(item_idx) + '.png')
plt.show()
pass
def PacfPloter(Data: pd.DataFrame):
for item_idx, item in Data.iteritems():
# 绘制偏自相关图
plot_pacf(item, lags=10).show()
plt.savefig('data/pics/' + '/PacfPlot_' + str(item_idx) + '.png')
plt.show()
pass
def ADFGetter(Data: pd.DataFrame, maxlags=5):
Data.dropna(axis=0, inplace=True)
adfResult = stm.tsa.stattools.adfuller(Data, maxlags)
output = pd.DataFrame(index=['Test Statistic Value', "p-value", "Lags Used", "Number of Observations Used",
"Critical Value(1%)", "Critical Value(5%)", "Critical Value(10%)"],
columns=['value'])
output['value']['Test Statistic Value'] = adfResult[0]
output['value']['p-value'] = adfResult[1]
output['value']['Lags Used'] = adfResult[2]
output['value']['Number of Observations Used'] = adfResult[3]
output['value']['Critical Value(1%)'] = adfResult[4]['1%']
output['value']['Critical Value(5%)'] = adfResult[4]['5%']
output['value']['Critical Value(10%)'] = adfResult[4]['10%']
return output
def PackedADFGetter(Data: pd.DataFrame):
result = []
for item_idx, item in Data.iteritems():
out = ADFGetter(pd.DataFrame(item))
result.append(out)
return Data, result
def CointDivide(Data: pd.DataFrame):
Data.dropna(axis=0, inplace=True)
result = []
for item_idx, item in Data.iteritems():
for item_idx_inner, item_inner in Data.iteritems():
if item_idx_inner != item_idx:
try:
out = stm.tsa.stattools.coint(item, item_inner)
# print(out)
result.append(out)
except Warning:
Data.drop([item_idx, item_idx_inner], axis=1)
return Data, result
def MAXVARFitter(Data: pd.DataFrame, varLagNum=5):
# 建立对象，dataframe就是前面的data，varLagNum就是你自己定的滞后阶数
orgMod = stm.tsa.VARMAX(Data, order=(varLagNum, 0), trend='n', exog=None)
# 估计：就是模型
fitMod = orgMod.fit(maxiter=1000, disp=False)
# 打印统计结果
# print(fitMod.summary())
# 获得模型残差
resid = fitMod.resid
return fitMod, resid
def VARFitter(Data: pd.DataFrame, lags=3):
VARModel = stm.tsa.VAR(Data)
result = VARModel.fit(lags)
# print(result.summary())
return VARModel, result
def FevdPic(FitMod, Name: str, file):
fevd = FitMod.fevd(10)
# 打印出方差分解的结果
print(fevd.summary(), file=file)
# 画图
fevd.plot(figsize=(12, 16))
plt.savefig('data/pics/' + '/FevdPic_' + Name + '.png')
plt.show()
pass
def CUSUM(resid):
# 原假设：无漂移（平稳），备择假设：有漂移（不平稳）
# het_breuschpagan()
# diagnostic
result = dia.breaks_cusumolsresid(resid)
return result
def ImpulsePic(fitMod, terms=3):
# orthogonalized=True，代表采用乔里斯基正交
ax = fitMod.impulse_responses(terms, orthogonalized=True).plot(figsize=(12, 16))
plt.savefig('data/pics/' + '/ImpulsePic' + '.png')
plt.show()
return None
def main():
plotFlag = False
Data_base, className = GetClasses()
# 不想再进行分析的数据，进行跳过，这里可以是空的
usedName = ['All Products', 'Mach and Elec', 'Vegetable',
'Transportation', 'Textiles and Clothing', 'Stone and Glass',
'Plastic or Rubber', 'Miscellaneous', 'Minerals', 'Metals',
'Hides and Skins', 'Capital goods', 'Fuels', 'Footwear', 'Food Products',
'Chemicals', 'Animal', 'Raw materials', 'Intermediate goods',
'Consumer goods', 'Wood']
for name in className:
print(name)
if name.strip() in usedName:
continue
input("any to init:")
Data = ReadTariffData(Data_base, name)
Data = DataProcesser(Data)
name = name.strip()
if plotFlag:
try:
file = open('data/' + name + '/' + name + '.txt', mode='w')
except FileNotFoundError:
file = open('data/' + name + '.txt', mode='w')
else:
file = None
if plotFlag:
TimeSeriesPloter(Data=Data, name=name)
input("any to start next: ")
AcfPloter(Data=Data)
input("any to start next: ")
PacfPloter(Data=Data)
info = PackedADFGetter(Data)[1]
info = CointDivide(Data)[1]
lags = 3
obj, fitMod_var = VARFitter(Data, lags)
print(fitMod_var.summary(), file=file)
if plotFlag:
FevdPic(fitMod_var, name, file=file)
input("anything to start maxvar: ")
fitMod_maxvar, resid = MAXVARFitter(Data, lags)
print(fitMod_maxvar.summary(), file=file)
input("any to start next: ")
if plotFlag:
ImpulsePic(fitMod_maxvar)
input("any to start next: ")
try:
if plotFlag:
CUSUM(resid)
except:
warnings.warn("CUSUM fail")
if file:
file.close()
onceMode = input("\nnext item or not? input num>0 to next:")
if int(onceMode) > 0:
continue
else:
break
pass
def afterMain():
pass
if __name__ == '__main__':
main()
数据的话是用的WITS的数据，就到这里啦