Adding scikit-learn (sklearn) prediction to pandas data frame

Question

I am trying to add a sklearn prediction to a pandas dataframe, so that I can make a thorough evaluation of the prediction. The relavant piece of code is the following:

clf = linear_model.LinearRegression()
clf.fit(Xtrain,ytrain)
ypred = pd.DataFrame({'pred_lin_regr': pd.Series(clf.predict(Xtest))})

The dataframes look like this:

Xtest

       axial_MET  cos_theta_r1  deltaE_abs  lep1_eta   lep1_pT  lep2_eta  
8000   1.383026      0.332365    1.061852  0.184027  0.621598 -0.316297   
8001  -1.054412      0.046317    1.461788 -1.141486  0.488133  1.011445   
8002   0.259077      0.429920    0.769219  0.631206  0.353469  1.027781   
8003  -0.096647      0.066200    0.411222 -0.867441  0.856115 -1.357888   
8004   0.145412      0.371409    1.111035  1.374081  0.485231  0.900024   

ytest

8000    1
8001    0
8002    0
8003    0
8004    0

ypred

        pred_lin_regr
0       0.461636
1       0.314448
2       0.363751
3       0.291858
4       0.416056

Concatenating Xtest and ytest works fine:

df_total = pd.concat([Xtest, ytest], axis=1)

but the event information is lost on ypred.

What would be the must python/pandas/numpy-like way to do this?

I am using the following versions:

argparse==1.2.1
cycler==0.9.0
decorator==4.0.4
ipython==4.0.0
ipython-genutils==0.1.0
matplotlib==1.5.0
nose==1.3.7
numpy==1.10.1
pandas==0.17.0
path.py==8.1.2
pexpect==4.0.1
pickleshare==0.5
ptyprocess==0.5
py==1.4.30
pyparsing==2.0.5
pytest==2.8.2
python-dateutil==2.4.2
pytz==2015.7
scikit-learn==0.16.1
scipy==0.16.1
simplegeneric==0.8.1
six==1.10.0
sklearn==0.0
traitlets==4.0.0
wsgiref==0.1.2

I tried the following:

df_total["pred_lin_regr"] = clf.predict(Xtest) 

seems to do the job, but I think I can't be sure that the events are matched correctly

Solution

You're correct with your second line, df_total["pred_lin_regr"] = clf.predict(Xtest) and it's more efficient.

In that one you're taking the output of clf.predict(), which happens to be an array, and adding it to a dataframe. The output you're receiving from the array itself is in order to match Xtest, since that's the case, adding it to a numpy array will not change or alter that order.

Here's a little proof from this example:

Taking the following protion:

import numpy as np

import pandas as pd
from sklearn import datasets, linear_model

# Load the diabetes dataset
diabetes = datasets.load_diabetes()

# Use only one feature
diabetes_X = diabetes.data[:, np.newaxis, 2]

# Split the data into training/testing sets
diabetes_X_train = diabetes_X[:-20]
diabetes_X_test = diabetes_X[-20:]

# Split the targets into training/testing sets
diabetes_y_train = diabetes.target[:-20]
diabetes_y_test = diabetes.target[-20:]

# Create linear regression object
regr = linear_model.LinearRegression()

# Train the model using the training sets
regr.fit(diabetes_X_train, diabetes_y_train)

print(regr.predict(diabetes_X_test))

df = pd.DataFrame(regr.predict(diabetes_X_test))

print(df)

The first print() function will give us a numpy array as expected:

[ 225.9732401   115.74763374  163.27610621  114.73638965  120.80385422
  158.21988574  236.08568105  121.81509832   99.56772822  123.83758651
  204.73711411   96.53399594  154.17490936  130.91629517   83.3878227
  171.36605897  137.99500384  137.99500384  189.56845268   84.3990668 ]

That order is identical to the second print() function in which we add the results to a dataframe:

             0
0   225.973240
1   115.747634
2   163.276106
3   114.736390
4   120.803854
5   158.219886
6   236.085681
7   121.815098
8    99.567728
9   123.837587
10  204.737114
11   96.533996
12  154.174909
13  130.916295
14   83.387823
15  171.366059
16  137.995004
17  137.995004
18  189.568453
19   84.399067

Rerunning the code for a portion of the test, will give us the same ordered results as such:

print(regr.predict(diabetes_X_test[0:5]))

df = pd.DataFrame(regr.predict(diabetes_X_test[0:5]))

print(df)

[ 225.9732401   115.74763374  163.27610621  114.73638965  120.80385422]
            0
0  225.973240
1  115.747634
2  163.276106
3  114.736390
4  120.803854

Comments

[asmgx]

How to fix the same problem if Xtest was a random selection and indexes were randomized, in that case we cannot match each record in the both dataframes