aspera1631

# gridsearch to find best regularization parameter
C_vect = np.logspace(-8, 5, 50)

# set the X and Y values.
X_tr = train_x[:1000,:]
Y_tr = train_y[:1000]
X_val = val_x[:1000,:]
Y_val = val_y[:1000]

aspera1631

# L1 regularization for C = 0.003
X = train_x[1:1000,:]
Y = train_y[1:1000]

# Specify the classifier. L1 penalty.
logreg_l1 = linear_model.LogisticRegression(penalty='l1', C=.003)
# train the classifier
logreg_l1.fit(X, Y)

# get the coefficients

aspera1631

# gridsearch to find best regularization parameter
C_vect = np.logspace(-8, 5, 50)

# set the X and Y values.
X_tr = train_x[:1000,:]
Y_tr = train_y[:1000]
X_val = val_x[:1000,:]
Y_val = val_y[:1000]

aspera1631

Classification report for classifier LogisticRegression(C=100000.0, class_weight=None, dual=False,
          fit_intercept=True, intercept_scaling=1, max_iter=100,
          multi_class='ovr', penalty='l2', random_state=None,
          solver='liblinear', tol=0.0001, verbose=0):
             precision    recall  f1-score   support

          0       0.93      0.97      0.95       634
          1       0.94      0.95      0.94       701
          2       0.92      0.86      0.89       628
          3       0.89      0.87      0.88       675

aspera1631

# Predict digits for the validation set
predicted = logreg.predict(val_x)
conf_matrix = metrics.confusion_matrix(val_y , predicted)

print("Classification report for classifier %s:\n%s\n"
      % (logreg, metrics.classification_report(val_y, predicted)))
print("Confusion matrix:\n%s" % conf_matrix)

# Let's plot the confusion matrix, but get rid of the diagonal to better show where the problems are.
off_diag = conf_matrix*(1 - np.identity(10))

aspera1631

# for the classifier, set the X and Y values.
X = train_x
Y = train_y


# Set regularization parameter (C)
reg_param = 1e5
# Define the classifier
logreg = linear_model.LogisticRegression(C=reg_param)
# train the classifier

aspera1631

# Code adapted from http://scikit-learn.org/stable/auto_examples/linear_model/plot_iris_logistic.html
# originally due to Gaël Varoquaux
# Modified by Brad Deutsch
# License: BSD 3 clause

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import matplotlib.image as mpimg

aspera1631

Holy Nova 
Blackwing Technician 
Dark Cultist 
Blackwing Corruptor 
The Coin 
.
.
.
Power Word: Shield 
Blackwing Corruptor 

aspera1631

                            Name
Entity ID                       
1                            NaN
2                            NaN
3                            NaN
4                   Anduin Wrynn
5                    Lesser Heal
6              Twilight Guardian
.
.

aspera1631

                            Name
Entity ID                       
1                            NaN
2                            NaN
3                            NaN
4                   Anduin Wrynn
5                    Lesser Heal
6              Twilight Guardian
7                            NaN
8                            NaN