notebook.community

Edit and run



In [1]:

    
import load_data, plots
import pandas as pd
from matplotlib import pyplot as plt
import statsmodels.api as sm



In [2]:

    
data= load_data.load_data()
plots.plot_data(data)



In [10]:

    
value = data['usd-blue-lanacion'].keys()[0]



In [11]:

    
value?



In [3]:

    
df = pd.DataFrame(data)



In [19]:

    
df['usd-bna/buy'].describe()









    Out[19]:





count    2613.000000
mean        3.575617
std         0.698774
min         2.780000
25%         3.020000
50%         3.180000
75%         4.000000
max         5.705000
dtype: float64



In [20]:

    
df['usd-bna/buy'].plot()









    Out[20]:





<matplotlib.axes.AxesSubplot at 0x106fb7d90>



In [21]:

    
df['usd-bna/buy'].hist()









    Out[21]:





<matplotlib.axes.AxesSubplot at 0x106fe5090>



In [22]:

    
df['usd-bna/buy'].apply(np.log10).hist()









    Out[22]:





<matplotlib.axes.AxesSubplot at 0x107056510>



In [4]:

    
serie = df['usd-bna/buy']



In [31]:

    
pd.stats.moments.rolling_mean(serie, 1).plot()









    Out[31]:





<matplotlib.axes.AxesSubplot at 0x10739ec10>



In [33]:

    
pd.stats.moments.rolling_mean(serie.apply(np.log10), 2).plot()









    Out[33]:





<matplotlib.axes.AxesSubplot at 0x10781dd10>



In [35]:

    
(serie/serie.shift(1)).plot()









    Out[35]:





<matplotlib.axes.AxesSubplot at 0x1078cf210>



In [50]:

    
(serie/serie.shift(1)).hist(bins=50)









    Out[50]:





<matplotlib.axes.AxesSubplot at 0x108ee49d0>



In [41]:

    
(serie/serie.shift(30)).plot()









    Out[41]:





<matplotlib.axes.AxesSubplot at 0x107c5b350>



In [42]:

    
(serie/serie.shift(360)).plot()









    Out[42]:





<matplotlib.axes.AxesSubplot at 0x107eac310>



In [49]:

    
(serie/serie.shift(30)).hist(bins=50)









    Out[49]:





<matplotlib.axes.AxesSubplot at 0x108e21990>



In [46]:

    
(serie/serie.shift(5)).plot()









    Out[46]:





<matplotlib.axes.AxesSubplot at 0x1085ccc10>



In [54]:

    
fig = (serie/serie.shift(5)).hist(bins=50)
fig.set_title('a')









    Out[54]:





<matplotlib.text.Text at 0x109b71c90>



In [64]:

    
for i in [5, 30, 360]:
    f, (ax1, ax2) = plt.subplots(2)
    ax1.hist((serie/serie.shift(i)).dropna(), bins=50)
    ax1.set_title('cada %d dias' % i)
    ax2.plot((serie/serie.shift(i)).dropna())
    ax2.set_title('cada %d dias' % i)
    f.savefig('plot-%d.pdf' % i, format = 'pdf')



In [57]:

    
plt.plot((serie/serie.shift(i)).dropna())









    Out[57]:





[<matplotlib.lines.Line2D at 0x10b080050>]



In [67]:

    
for i in [5, 30, 360]:
    f, (ax1, ax2) = plt.subplots(2)
    ax1.hist((serie/serie.shift(i)).apply(np.log10).dropna(), bins=50)
    ax1.set_title('cada %d dias' % i)
    ax2.plot((serie/serie.shift(i)).apply(np.log10).dropna())
    ax2.set_title('cada %d dias' % i)



In [5]:

    
def difference(serie, shift):
    return serie/serie.shift(i)



In [69]:

    
for i in [5, 30, 360]:
    f, (ax1, ax2) = plt.subplots(2)
    ax1.hist(difference(serie, i).dropna(), bins=50)
    ax1.set_title('cada %d dias' % i)
    ax2.plot(difference(serie, i).dropna())
    ax2.set_title('cada %d dias' % i)



In [70]:

    
for i in [5, 30, 360]:
    f, (ax1, ax2) = plt.subplots(2)
    ax1.hist(difference(difference(serie, i), i).dropna(), bins=50)
    ax1.set_title('cada %d dias' % i)
    ax2.plot(difference(difference(serie, i), i).dropna())
    ax2.set_title('cada %d dias' % i)



In [7]:

    
logged = serie.apply(np.log10)



In [73]:

    
(logged - logged.shift(1)).plot()









    Out[73]:





<matplotlib.axes.AxesSubplot at 0x10b0f6810>



In [6]:

    
df1 = lambda x, t=1: x - x.shift(t)



In [82]:

    
df1(logged, 5).plot()









    Out[82]:





<matplotlib.axes.AxesSubplot at 0x10c36e050>



In [84]:

    
df1(df1(logged, 5), 5).plot()









    Out[84]:





<matplotlib.axes.AxesSubplot at 0x10caf0890>



In [79]:

    
df1(df1(df1(logged))).hist()









    Out[79]:





<matplotlib.axes.AxesSubplot at 0x10b52d150>



In [29]:

    
log_diff = df1(df1(logged, 5), 5)



In [32]:

    
arma_mod = sm.tsa.ARMA(log_diff.dropna(),order=(2,2), freq='D')



In [33]:

    
arma_res = arma_mod.fit()



In [99]:

    
arma_res.fittedvalues.plot()









    Out[99]:





<matplotlib.axes.AxesSubplot at 0x10ee58e50>



In [104]:

    
forecast = arma_res.forecast(100)[0]



In [108]:

    
arima = sm.tsa.ARMA(logged.dropna(), order = (2,2,2), freq='D')









    



---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-108-bda19b204028> in <module>()
----> 1 arima = sm.tsa.ARMA(logged.dropna(), order = (2,2,2), freq='D')

TypeError: __init__() got an unexpected keyword argument 'order'



In [111]:

    
df1(df1(logged, 30), 30).plot()









    Out[111]:





<matplotlib.axes.AxesSubplot at 0x10f9364d0>



In [17]:

    
sm.graphics.tsa.plot_acf(df1(df1(logged, 5), 5).dropna(), lags=10)









    Out[17]:



In [18]:

    
sm.graphics.tsa.plot_pacf(df1(df1(logged, 5), 5).dropna(), lags=10)









    Out[18]:



In [22]:

    
sm.graphics.tsa.plot_acf(df1(df1(df1(logged, 5), 5), 5).dropna(), lags=20)









    Out[22]:



In [26]:

    
sm.graphics.tsa.plot_acf(df1(logged, 5).dropna(), lags=20)









    Out[26]:



In [27]:

    
sm.graphics.tsa.plot_pacf(df1(logged, 5).dropna(), lags=20)









    Out[27]:



In [36]:

    
sm.graphics.tsa.plot_pacf(logged.dropna(), lags=20)









    Out[36]:



In [38]:

    
sm.graphics.tsa.plot_acf(logged.dropna(), lags=100)









    Out[38]:

Esto indicaría un modelo ARIMA(0,2,9)



In [46]:

    
sm.graphics.tsa.plot_acf(df1(df1(logged, 5), 5).dropna(), lags=30)
sm.graphics.tsa.plot_pacf(df1(df1(logged, 5), 5).dropna(), lags=30)









    Out[46]:



In [47]:

    
sm.graphics.tsa.plot_acf(logged.dropna(), lags=100)
sm.graphics.tsa.plot_pacf(logged.dropna(), lags=100)









    Out[47]:



In [51]:

    
df1(df1(logged, 5), 5).hist(bins=40)









    Out[51]:





<matplotlib.axes.AxesSubplot at 0x10c0f1c10>



In [49]:

    
df1(df1(logged, 5), 5).plot()









    Out[49]:





<matplotlib.axes.AxesSubplot at 0x10c074ad0>



In [ ]: