In [1]:

    

from harness import Harness

import sklearn.datasets
import sklearn.discriminant_analysis
import sklearn.mixture
import sklearn.model_selection

from pandas import ( 
    CategoricalIndex, DataFrame, Index, Series
)

from IPython.display import (
    Markdown,
)


    

In [2]:

    

"""Can I load the iris data"""
iris = sklearn.datasets.load_iris()


    

In [3]:

    

"""Can I initialize a Harness DataFrame"""
df = Harness(
    data=iris['data'], 
    index=CategoricalIndex(iris['target']).rename('target'),
    columns=iris['feature_names'],
    estimator=sklearn.discriminant_analysis.LinearDiscriminantAnalysis(),
    feature_level='target',
)

df = df.set_index(
    df.index
    .rename_categories(iris['target_names'])
    .rename('target_name'), append=True
).set_index(
    df
    .index.rename_categories(['red', 'green', 'blue'])
    .rename('color'), append=True
)

df.sample(5)


    

    
Out[3]:






  

    

      

      

      

      
sepal length (cm)
      
sepal width (cm)
      
petal length (cm)
      
petal width (cm)
    
    

      
target
      
target_name
      
color
      

      

      

      

    
  
  

    

      
2
      
virginica
      
blue
      
6.4
      
2.8
      
5.6
      
2.1
    
    

      
blue
      
6.9
      
3.2
      
5.7
      
2.3
    
    

      
blue
      
6.7
      
3.3
      
5.7
      
2.5
    
    

      
blue
      
5.7
      
2.5
      
5.0
      
2.0
    
    

      
0
      
setosa
      
red
      
5.4
      
3.9
      
1.3
      
0.4
    
  

In [4]:

    

test_train = next(
    sklearn.model_selection
    .StratifiedKFold(n_splits=2)
    .split(df.values, df.Index('target'))
)

split = DataFrame(index=df.index, columns=['split'])
split.iloc[test_train[0]], split.iloc[test_train[1]] = 'train', 'test'
df = df.set_index(split.set_index('split', append=True).swaplevel(-1,0).index)

df.ix['train'].fit().sample(5)


    

    
Out[4]:






  

    

      

      

      

      
sepal length (cm)
      
sepal width (cm)
      
petal length (cm)
      
petal width (cm)
    
    

      
target_name
      
color
      
target
      

      

      

      

    
  
  

    

      
setosa
      
red
      
0
      
4.7
      
3.2
      
1.6
      
0.2
    
    

      
0
      
5.0
      
3.3
      
1.4
      
0.2
    
    

      
versicolor
      
green
      
1
      
6.7
      
3.0
      
5.0
      
1.7
    
    

      
1
      
5.7
      
2.9
      
4.2
      
1.3
    
    

      
virginica
      
blue
      
2
      
6.9
      
3.1
      
5.4
      
2.1
    
  

In [5]:

    

from sklearn.metrics import confusion_matrix


    

In [6]:

    

from sklearn import preprocessing, decomposition, discriminant_analysis


    

In [7]:

    

df = (
    df
    .set_params(estimator=preprocessing.RobustScaler())
    .fit_transform()
    .set_params(estimator=decomposition.PCA())
    .fit_transform()
    .set_params(estimator=discriminant_analysis.LinearDiscriminantAnalysis())
)
df.ix['train'].fit().sample(2)


    

    
Out[7]:






  

    

      

      

      

      
0
      
1
      
2
      
3
    
    

      
target_name
      
color
      
target
      

      

      

      

    
  
  

    

      
setosa
      
red
      
0
      
1.436901
      
-0.033947
      
-0.002798
      
-0.055323
    
    

      
virginica
      
blue
      
2
      
-1.228911
      
-0.415372
      
-0.009152
      
-0.112343
    
  

In [8]:

    

(
    df
    .transform()
    .set_index(
        CategoricalIndex(df.predict()[0], name='prediction'), append=True
    )
    .sample(2)
)


    

    
Out[8]:






  

    

      

      

      

      

      

      
0
      
1
    
    

      
split
      
target_name
      
color
      
target
      
prediction
      

      

    
  
  

    

      
test
      
versicolor
      
green
      
1
      
1
      
-1.184620
      
-1.742177
    
    

      
train
      
setosa
      
red
      
0
      
0
      
7.017018
      
0.623815
    
  

In [9]:

    

df.ix['train'].score()


    

    
Out[9]:





0.97333333333333338

In [10]:

    

df.add_template(
scoreboard="""
The table presents some information about the `{{
    df.estimator.__str__().split('(',1)[0]
}}` model applied to the `iris` dataset.

|         |           Test          |           Training       |
|---------|-------------------------|--------------------------|
| Score   |{{df.ix['test'].score()}}|{{df.ix['train'].score()}}|
| Samples | {{df.ix['test'] | len}} | {{df.ix['train'] | len}} |
"""
);


    

In [11]:

    

Markdown(df.get_template('scoreboard'))


    

    
Out[11]:




The table presents some information about the LinearDiscriminantAnalysis model applied to the iris dataset.



Test
Training




Score
0.986666666667
0.973333333333


Samples
75
75

In [12]:

    

transformed = df.transform()
transformed.sample(2)


    

    
Out[12]:






  

    

      

      

      

      

      
0
      
1
    
    

      
split
      
target_name
      
color
      
target
      

      

    
  
  

    

      
train
      
virginica
      
blue
      
2
      
-5.364102
      
-0.009331
    
    

      
setosa
      
red
      
0
      
7.341003
      
-0.791289
    
  

In [13]:

    

with transformed.reset_index().DataSource(
    x=0, y=1, fill_color='color', text='target_name'
) as source:
    source.Scatter(color='text', marker='split')
    source.save()