Add code to experiement with param values

Author: maillouxc

part2/experiments.py

@@ -11,42 +11,26 @@
 from sklearn import svm
 from sklearn.linear_model import Perceptron
 
+training_emails = []
+training_labels = []
+test_emails = []
+test_labels = []
 
 def main():
- parsed_args = parse_command_line_arguments()
- 
+ global training_emails, training_labels, test_emails, test_labels
+ parsed_args = parse_command_line_arguments() 
  data_path = parsed_args.path
  training_emails, training_labels, = load_dataset(data_path + "/training")
  test_emails, test_labels = load_dataset(data_path + "/test")
 
- svm_classifier = Pipeline([
- ('count_vectorizer', CountVectorizer(stop_words="english")),
- ('tfidf_transformer', TfidfTransformer()),
- ('svm_classifier', svm.SVC(gamma="scale",
- C=1,
- verbose=False))
- ])
-
- # Train the svm
- svm_classifier.fit(training_emails, training_labels)
-
- num_correctly_classified = 0
- num_wrongly_classified = 0
- # Test it
- results = svm_classifier.predict(test_emails)
- for i in range(len(results)):
- if (results[i] == test_labels[i]):
- num_correctly_classified += 1
- else:
- num_wrongly_classified += 1
-
- accuracy = str("{0:.3%}").format(numpy.mean(results == test_labels))
- print("Accuracy of SVM classifier was " + accuracy)
+ test_svm_classifier(C=1, use_idf=True, kernel='linear')
+ test_svm_classifier(C=2, use_idf=True, kernel='linear')
+ test_svm_classifier(C=10, use_idf=True, kernel='linear')
 
 
 def parse_command_line_arguments():
  parser = argparse.ArgumentParser(
- description="Classifies emails as spam or ham")
+ description="Tests two types of email spam classifiers")
  parser.add_argument(
  'path',
  help="The path to the two folders, \'test\' and \'training\'"
@@ -74,6 +58,36 @@ def load_dataset(data_path):
  return email_data, email_labels
 
 
+def test_svm_classifier(C, use_idf, kernel):
+ svm_classifier = Pipeline([
+ ('count_vectorizer', CountVectorizer(stop_words="english")),
+ ('tfidf_transformer', TfidfTransformer(use_idf=use_idf)),
+ ('svm_classifier', svm.SVC(gamma="scale",
+ C=C,
+ class_weight=None,
+ kernel=kernel
+ verbose=False))
+ ])
+
+ print(svm_classifier.get_params()['svm_classifier'])
+
+ # Train the svm
+ svm_classifier.fit(training_emails, training_labels)
+
+ # Test it
+ num_correctly_classified = 0
+ num_wrongly_classified = 0
+ results = svm_classifier.predict(test_emails)
+ for i in range(len(results)):
+ if (results[i] == test_labels[i]):
+ num_correctly_classified += 1
+ else:
+ num_wrongly_classified += 1
+
+ accuracy = str("{0:.3%}").format(numpy.mean(results == test_labels))
+ print("Accuracy of SVM classifier was " + accuracy)
+
+
 if __name__ == "__main__":
  main()