In [1]:

    

%load_ext sql
%matplotlib inline


    

In [2]:

    

import logging
import os
import sys

from configparser import ConfigParser
import pandas as pd
import numpy as np
import matplotlib
import matplotlib.pyplot as plt


    

In [3]:

    

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
config = ConfigParser()
configfn = os.path.join(os.path.expanduser('~'), '.pgcred')
config.read(configfn)


    

    
Out[3]:





['/home/sansbacon/.nbadb']

In [4]:

    

user=config['nbadb']['username']
password=config['nbadb']['password']
db=config['nbadb']['database']


    

In [5]:

    

connection_string = "postgresql://{user}:{password}@localhost/{db}".format(user=user, password=password, db=db)
%sql $connection_string


    

    
Out[5]:





u'Connected: nbadb@nbadb'

In [25]:

    

from tpot import TPOTClassifier
from sklearn.model_selection import train_test_split
from sklearn import preprocessing


    

In [55]:

    

result = %sql SELECT * FROM tmpmodel


    

    




5586 rows affected.

In [56]:

    

df = result.DataFrame()
df.head(10)


    

    
Out[56]:






  

    

      

      
salary
      
minema
      
dkema
      
consensus_game_ou
      
dtot
      
back_to_back
      
three_in_four
      
def_rating_ema
      
pace_ema
      
y
    
  
  

    

      
0
      
5000
      
23.3
      
22.8
      
197.0
      
-4.44
      
0
      
0
      
98.2
      
98.3
      
0
    
    

      
1
      
5200
      
25.0
      
22.6
      
196.5
      
-7.08
      
0
      
0
      
107.8
      
93.0
      
1
    
    

      
2
      
5800
      
26.5
      
27.2
      
207.5
      
-1.12
      
0
      
0
      
110.6
      
96.0
      
1
    
    

      
3
      
6000
      
27.4
      
28.0
      
198.25
      
-4.65
      
0
      
0
      
97.1
      
97.2
      
1
    
    

      
4
      
6100
      
27.5
      
29.6
      
197.5
      
-4.98
      
0
      
0
      
106.1
      
98.1
      
0
    
    

      
5
      
5900
      
27.4
      
29.5
      
200.5
      
0.75
      
0
      
0
      
105.7
      
101.6
      
1
    
    

      
6
      
5400
      
27.8
      
29.6
      
188.5
      
-7.69
      
1
      
1
      
101.9
      
95.4
      
0
    
    

      
7
      
5500
      
27.1
      
26.1
      
207.5
      
2.06
      
0
      
1
      
106.2
      
102.1
      
0
    
    

      
8
      
5500
      
25.3
      
25.8
      
201.0
      
-3.61
      
0
      
0
      
107.3
      
98.4
      
1
    
    

      
9
      
5000
      
25.2
      
27.6
      
195.5
      
-15.79
      
0
      
0
      
109.1
      
99.4
      
1
    
  

In [57]:

    

df = df.dropna(how='any')


    

In [58]:

    

X_train, X_test, y_train, y_test = train_test_split(df.ix[:,:-1], df.ix[:,-1],
                                                    train_size=0.75, test_size=0.25)


    

In [59]:

    

X_train_scale = preprocessing.scale(X_train)
X_test_scale = preprocessing.scale(X_test)


    

In [60]:

    

tpot = TPOTClassifier(generations=5, population_size=20, verbosity=2)
tpot.fit(X_train_scale, y_train)
print(tpot.score(X_test_scale, y_test))


    

    




Optimization Progress:  17%|█▋        | 20/120 [00:35<02:20,  1.40s/pipeline]





    




Version 0.6.8 of tpot is outdated. Version 0.7.0 was released Wednesday March 22, 2017.
Generation 1 - Current best internal CV score: 0.531110408573





    




Optimization Progress:  32%|███▏      | 38/120 [03:28<12:16,  8.98s/pipeline]





    




Generation 2 - Current best internal CV score: 0.531110408573





    




Optimization Progress:  49%|████▉     | 59/120 [06:11<11:07, 10.95s/pipeline]





    




Generation 3 - Current best internal CV score: 0.540038846303





    




Optimization Progress:  62%|██████▎   | 75/120 [08:07<05:38,  7.51s/pipeline]





    




Generation 4 - Current best internal CV score: 0.540038846303





    




Optimization Progress:  82%|████████▏ | 98/120 [10:36<03:20,  9.11s/pipeline]





    




Generation 5 - Current best internal CV score: 0.540038846303





    




                                                                              





    





Best pipeline: ExtraTreesClassifier(input_matrix, 43, 0.82000000000000006)
0.512391465097






    

In [61]:

    

from sklearn.ensemble import ExtraTreesClassifier

forest = ExtraTreesClassifier(n_estimators=250, random_state=0)


    

In [62]:

    

forest.fit(X_train_scale, y_train)


    

    
Out[62]:





ExtraTreesClassifier(bootstrap=False, class_weight=None, criterion='gini',
           max_depth=None, max_features='auto', max_leaf_nodes=None,
           min_impurity_split=1e-07, min_samples_leaf=1,
           min_samples_split=2, min_weight_fraction_leaf=0.0,
           n_estimators=250, n_jobs=1, oob_score=False, random_state=0,
           verbose=0, warm_start=False)

In [63]:

    

importances = forest.feature_importances_
std = np.std([tree.feature_importances_ for tree in forest.estimators_],
             axis=0)
indices = np.argsort(importances)[::-1]


    

In [66]:

    

print("Feature ranking:")

for f in range(X_train_scale.shape[1]):
    print("%d. feature %d (%f)" % (f + 1, indices[f], importances[indices[f]]))

plt.figure()
plt.title("Feature importances")
plt.bar(range(X_train_scale.shape[1]), importances[indices],
       color="r", yerr=std[indices], align="center")
plt.xticks(range(X_train_scale.shape[1]), indices)
plt.xlim([-1, X_train_scale.shape[1]])
plt.show()


    

    




Feature ranking:
1. feature 2 (0.159057)
2. feature 1 (0.158966)
3. feature 0 (0.148447)
4. feature 4 (0.131073)
5. feature 7 (0.128373)
6. feature 8 (0.127443)
7. feature 3 (0.126365)
8. feature 6 (0.010289)
9. feature 5 (0.009987)






    

In [68]:

    

df.columns


    

    
Out[68]:





Index([u'salary', u'minema', u'dkema', u'consensus_game_ou', u'dtot', u'back_to_back', u'three_in_four', u'def_rating_ema', u'pace_ema', u'y'], dtype='object')

In [ ]: