In [2]:

    

import os
import math
import pandas as pd
import numpy as np
from scipy import stats
# print stats.hmean([ -50.2 , 100.5 ])


    

In [52]:

    

path = '/home/zongyi/bimbo_data/submission/'
for f in os.listdir(path):
    print f


    

    




w9_f14_3xn500_c0.7_mean_45482.csv
w9_n1000_45133.csv
w9_n500_45423.csv
w8_f14_n1000_c0.7_45736.csv
ensemble-kele-0442.csv
w9_full1200_47233.csv
w8_f15_n1000_c0.7_mean_46121.csv
w8_f15_n1000_c0.7_48181.csv
w9_n500_45282.csv
w9_f25_n1500_c0.5_46191.csv
w9_full1.csv
w8_n500_45918.csv
w9_f14_2n1000_c0.7_mean_4539.csv
w9_f17_1500_49xx.csv
ensemble_full_04387.csv
w9_full_47401.csv
w8_f14_n1200_c0.7.3_47317.csv
w8_n500_45986.csv
w8_n1000_48381.csv
ensemble-452-474-45034.csv
w9_f17_n500_c0.6_48153.csv
w8_f26_n1000_c0.5_46928.csv

In [69]:

    

df1 = pd.read_csv(path+'w9_n1000_45133.csv')
df2 = pd.read_csv(path+'w9_f25_n1500_c0.5_46191.csv')

df3 = pd.read_csv(path+'w9_full1200_47233.csv')
df4 = pd.read_csv(path+'w9_f14_2n1000_c0.7_mean_4539.csv')

df5 = pd.read_csv(path+'w9_f14_3xn500_c0.7_mean_45482.csv')


    

In [70]:

    

df1.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd1'}, inplace=True)
df2.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd2'}, inplace=True)
df3.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd3'}, inplace=True)
df4.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd4'}, inplace=True)
df5.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd5'}, inplace=True)


    

In [71]:

    

df = pd.merge(df1, df2, how='left', on=['id'])
df = pd.merge(df, df3, how='left', on=['id'])
df = pd.merge(df, df4, how='left', on=['id'])
df = pd.merge(df, df5, how='left', on=['id'])


    

In [74]:

    

print stats.pearsonr(df['d1'], df['d2'])
print stats.pearsonr(df['d1'], df['d3'])
print stats.pearsonr(df['d1'], df['d4'])
print stats.pearsonr(df['d1'], df['d5'])
print stats.pearsonr(df['d2'], df['d3'])
print stats.pearsonr(df['d2'], df['d4'])
print stats.pearsonr(df['d2'], df['d5'])
print stats.pearsonr(df['d3'], df['d4'])
print stats.pearsonr(df['d3'], df['d5'])
print stats.pearsonr(df['d4'], df['d5'])


    

    




(0.90859839944163689, 0.0)
(0.92805488735048192, 0.0)
(0.88624683036048535, 0.0)
(0.87027126832752355, 0.0)
(0.92529220805822276, 0.0)
(0.92288602605536907, 0.0)
(0.90392357835496917, 0.0)
(0.91427252279179827, 0.0)
(0.89463221754723543, 0.0)
(0.93983579944164275, 0.0)

In [77]:

    

df[:3]


    

    
Out[77]:






  

    

      

      
d1
      
id
      
d2
      
d3
      
d4
      
d5
      
Demanda_uni_equil
    
  
  

    

      
0
      
3.32189
      
4721633
      
3.43315
      
3.22524
      
3.24967
      
3.23493
      
3.315742
    
    

      
1
      
2.08302
      
6035419
      
2.29651
      
2.08906
      
2.01552
      
1.89152
      
2.093581
    
    

      
2
      
1.62305
      
3462602
      
1.70056
      
1.65728
      
1.52238
      
1.61837
      
1.627146
    
  

In [ ]:

    

def f(x):
    return 0.3 * (0.7 * x['d1'] + 0.3 * x['d2']) + 0.7 * x['d3']
df['Demanda_uni_equil'] = df[['d1', 'd2','d3','d4','d5']].apply(f, axis=1)


    

In [84]:

    

df['Demanda_uni_equil'] = 0.5*df['d1'] + 0.25*(0.55*df['d4']+0.45*df['d5']) + 0.25*(0.65*df['d2']+0.35*df['d3'])


    

In [ ]:

    




    

In [ ]:

    




    

In [78]:

    

path = '/home/zongyi/bimbo_data/submission/'
for f in os.listdir(path):
    print f


    

    




w9_f14_3xn500_c0.7_mean_45482.csv
w9_n1000_45133.csv
w9_n500_45423.csv
w8_f14_n1000_c0.7_45736.csv
ensemble-kele-0442.csv
w9_full1200_47233.csv
w8_f15_n1000_c0.7_mean_46121.csv
w8_f15_n1000_c0.7_48181.csv
w9_n500_45282.csv
w9_f25_n1500_c0.5_46191.csv
w9_full1.csv
w8_n500_45918.csv
w9_f14_2n1000_c0.7_mean_4539.csv
w9_f17_1500_49xx.csv
ensemble_full_04387.csv
w9_full_47401.csv
w8_f14_n1200_c0.7.3_47317.csv
w8_n500_45986.csv
w8_n1000_48381.csv
ensemble-452-474-45034.csv
w9_f17_n500_c0.6_48153.csv
w8_f26_n1000_c0.5_46928.csv

In [81]:

    

df6 = pd.read_csv(path+'w8_f14_n1000_c0.7_45736.csv')
df7 = pd.read_csv(path+'w8_f15_n1000_c0.7_mean_46121.csv')
df8 = pd.read_csv(path+'w8_f26_n1000_c0.5_46928.csv')


    

In [82]:

    

df6.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd6'}, inplace=True)
df7.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd7'}, inplace=True)
df8.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd8'}, inplace=True)
df = pd.merge(df, df6, how='left', on=['id'])
df = pd.merge(df, df7, how='left', on=['id'])
df = pd.merge(df, df8, how='left', on=['id'])


    

In [83]:

    

print stats.pearsonr(df['d6'], df['d7'])
print stats.pearsonr(df['d6'], df['d8'])
print stats.pearsonr(df['d7'], df['d8'])


    

    




(0.87384838745332227, 0.0)
(0.90843548653792394, 0.0)
(0.93853149179902695, 0.0)

In [85]:

    

df['Demanda_uni_equil1'] = 0.6*df['d6'] + 0.2*df['d7'] + 0.2*df['d8']


    

In [90]:

    

df[:3]


    

    
Out[90]:






  

    

      

      
d1
      
id
      
d2
      
d3
      
d4
      
d5
      
Demanda_uni_equil
      
d6
      
d7
      
d8
      
Demanda_uni_equil1
      
d10
    
  
  

    

      
0
      
3.32189
      
4721633
      
3.43315
      
3.22524
      
3.24967
      
3.23493
      
3.554216
      
3.75740
      
3.58134
      
3.59559
      
3.689826
      
3.615020
    
    

      
1
      
2.08302
      
6035419
      
2.29651
      
2.08906
      
2.01552
      
1.89152
      
2.055526
      
1.94513
      
1.81329
      
2.16537
      
1.962810
      
2.067495
    
    

      
2
      
1.62305
      
3462602
      
1.70056
      
1.65728
      
1.52238
      
1.61837
      
1.631257
      
1.71653
      
1.69398
      
1.75552
      
1.719818
      
1.610435
    
  

In [ ]:

    




    

In [87]:

    

df10 = pd.read_csv(path+'ensemble-kele-0442.csv')
df10.rename(columns={'id': 'id', 'Demanda_uni_equil': 'd10'}, inplace=True)
df = pd.merge(df, df10, how='left', on=['id'])


    

In [89]:

    

df['Demanda_uni_equil'] = 0.4*(0.6*df['Demanda_uni_equil'] + 0.4*df['Demanda_uni_equil1']) + 0.6*df['d10']


    

In [ ]:

    




    

In [18]:

    

# result['d1'] = result['d1'].apply(lambda x: max(x,0.000000001))    
# result['d2'] = result['d2'].apply(lambda x: max(x,0.000000001))
# result['d3'] = result['d3'].apply(lambda x: max(x,0.000000001))


    

In [19]:

    

def f(x):
    return stats.hmean(x)
result['Demanda_uni_equil'] = result[['d1', 'd2', 'd3']].apply(f, axis=1)


    

In [11]:

    

def f(x):
    return max(math.expm1(0.25 * math.log1p(x['d1']) + 0.5 * math.log1p(x['d2']) + 0.25 * math.log1p(x['d3'])),0)
result['Demanda_uni_equil'] = result[['d1', 'd2', 'd3']].apply(f, axis=1)


    

In [9]:

    

def f(x):
    return 0.1 * x['d1'] + 0.6 * x['d2'] + 0.3 * x['d3']
result['Demanda_uni_equil'] = result[['d1', 'd2', 'd3']].apply(f, axis=1)


    

In [11]:

    

def f(x):
    return 0.6 * x['d2'] + 0.4 * x['d3']
result['Demanda_uni_equil'] = result[['d2', 'd3']].apply(f, axis=1)


    

In [ ]:

    




    

In [91]:

    

sub = df[['id','Demanda_uni_equil']]
sub.to_csv(path+'ensemble_fulll.csv', index=False)