We import a bunch of libraries



In [1]:

    
import pandas as pd
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction import DictVectorizer
from sklearn.metrics import accuracy_score, precision_score, recall_score
from sklearn.neural_network import MLPClassifier









    



/Users/vikas/workspace/scikit-learn/sklearn/cross_validation.py:42: DeprecationWarning: This module has been deprecated in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.
  "This module will be removed in 0.20.", DeprecationWarning)

Load the data



In [2]:

    
countries = pd.read_csv("countries.csv", names=['name', 'continent'])

Configure feature extractions.



In [11]:

    
cv = CountVectorizer(analyzer='char', ngram_range=(1,1))
vec = DictVectorizer(sparse=False)

Transform inputs and outputs into a suitable format.



In [12]:

    
inputs = cv.fit_transform(countries.name)
outputs = vec.fit_transform(countries.continent.apply(lambda c: {'continent': c}))

Train a neural network.



In [62]:

    
model = MLPClassifier(activation='logistic', algorithm='l-bfgs', hidden_layer_sizes=(25))

Fit the entire input set. We don't care about overfitting. We want a 100% fit.



In [63]:

    
model.fit(inputs, outputs)
predicted_outputs = model.predict(inputs)

We compute the accuracy, precision and recall. All of these need to be 1.



In [64]:

    
accuracy = accuracy_score(outputs, predicted_outputs)
precision = precision_score(outputs, predicted_outputs, average='samples')
recall = recall_score(outputs, predicted_outputs, average='samples')

(accuracy, precision, recall)









    Out[64]:





(1.0, 1.0, 1.0)



In [65]:

    
def predict_continent(country):
    X = cv.transform([country])
    y = model.predict(X)

    for k, v in vec.inverse_transform(y)[0].items():
        if v == 1.0:
            return k.split('=')[1].title()
        
    return "Unknown"

This function can now predict the continent perfectly.



In [66]:

    
print(predict_continent("India"))
print(predict_continent("United States"))









    



Asia
North_America



In [67]:

    
[c.shape for c in model.coefs_]









    Out[67]:





[(32, 25), (25, 7)]