|
m@16
|
1 # -*- coding: utf-8 -*-
|
|
m@16
|
2 """
|
|
m@16
|
3 Created on Thu Nov 10 15:10:32 2016
|
|
m@16
|
4
|
|
m@16
|
5 @author: mariapanteli
|
|
m@16
|
6 """
|
|
m@16
|
7 import numpy as np
|
|
m@16
|
8 import pandas as pd
|
|
m@16
|
9 from sklearn import metrics
|
|
m@16
|
10
|
|
m@16
|
11 import map_and_average
|
|
m@16
|
12 import util_feature_learning
|
|
m@16
|
13
|
|
m@16
|
14
|
|
m@16
|
15 FILENAMES = map_and_average.OUTPUT_FILES
|
|
m@16
|
16
|
|
m@16
|
17
|
|
m@16
|
18 def load_data_from_pickle(filename):
|
|
m@16
|
19 X_list, Y, Yaudio = pickle.load(open(filename,'rb'))
|
|
m@16
|
20 X = np.concatenate(data_list, axis=1)
|
|
m@16
|
21 return X, Y, Yaudio
|
|
m@16
|
22
|
|
m@16
|
23
|
|
m@16
|
24 def get_train_test_indices():
|
|
m@16
|
25 trainset, valset, testset = map_and_average.load_train_val_test_sets()
|
|
m@16
|
26 trainaudiolabels, testaudiolabels = trainset[2], testset[2]
|
|
m@16
|
27 # train, test indices
|
|
m@16
|
28 aa_train = np.unique(trainaudiolabels)
|
|
m@16
|
29 aa_test = np.unique(testaudiolabels)
|
|
m@16
|
30 traininds = np.array([i for i, item in enumerate(audiolabs) if item in aa_train])
|
|
m@16
|
31 testinds = np.array([i for i, item in enumerate(audiolabs) if item in aa_test])
|
|
m@16
|
32 return traininds, testinds
|
|
m@16
|
33
|
|
m@16
|
34
|
|
m@16
|
35 def get_train_test_sets(X, Y, traininds, testinds):
|
|
m@16
|
36 X_train = X[traininds, :]
|
|
m@16
|
37 Y_train = Y[traininds]
|
|
m@16
|
38 X_test = X[testinds, :]
|
|
m@16
|
39 Y_test = Y[testinds]
|
|
m@16
|
40 return X_train, Y_train, X_test, Y_test
|
|
m@16
|
41
|
|
m@16
|
42
|
|
m@16
|
43 def classify_for_filenames(file_list=FILENAMES):
|
|
m@16
|
44 df_results = pd.DataFrame()
|
|
m@16
|
45 feat_learner = util_feature_learning.Transformer()
|
|
m@16
|
46 for filename in file_list:
|
|
m@16
|
47 X, Y, Yaudio = load_data_from_pickle(filename)
|
|
m@16
|
48 traininds, testinds = get_train_test_indices()
|
|
m@16
|
49 X_train, Y_train, X_test, Y_test = get_train_test_sets(X, Y, traininds, testinds)
|
|
m@16
|
50 df_result = feat_learner.classify(X_train, Y_train, X_test, Y_test)
|
|
m@16
|
51 df_results = pd.concat([df_results, df_result], axis=0, ignore_index=True)
|
|
m@16
|
52 return df_results
|
|
m@16
|
53
|
|
m@16
|
54
|
|
m@16
|
55 def plot_CF(CF, labels=None, figurename=None):
|
|
m@16
|
56 labels[labels=='United States of America'] = 'United States Amer.'
|
|
m@16
|
57 plt.imshow(CF, cmap="Greys")
|
|
m@16
|
58 plt.xticks(range(len(labels)), labels, rotation='vertical', fontsize=4)
|
|
m@16
|
59 plt.yticks(range(len(labels)), labels, fontsize=4)
|
|
m@16
|
60 if figurename is not None:
|
|
m@16
|
61 plt.savefig(figurename, bbox_inches='tight')
|
|
m@16
|
62
|
|
m@16
|
63
|
|
m@16
|
64 def confusion_matrix(X_train, Y_train, X_test, Y_test, saveCF=False, plots=False):
|
|
m@16
|
65 feat_learner = util_feature_learning.Transformer()
|
|
m@16
|
66 accuracy, predictions = util_feature_learning.classification_accuracy(X_train, Y_train,
|
|
m@16
|
67 X_test, Y_test, model=util_feature_learning.modelLDA)
|
|
m@16
|
68 labels = np.unique(Y_test) # TODO: countries in geographical proximity
|
|
m@16
|
69 CF = metrics.confusion_matrix(Y_test, predictions, labels=labels)
|
|
m@16
|
70 if saveCF:
|
|
m@16
|
71 np.savetxt('data/CFlabels.csv', labels, fmt='%s')
|
|
m@16
|
72 np.savetxt('data/CF.csv', CF, fmt='%10.5f')
|
|
m@16
|
73 if plots:
|
|
m@16
|
74 plot_CF(CF, labels=labels, figurename='data/conf_matrix.pdf')
|
|
m@16
|
75 return accuracy, predictions
|
|
m@16
|
76
|
|
m@16
|
77
|
|
m@16
|
78 if __name__ == '__main__':
|
|
m@16
|
79 df_results = classify_for_filenames(file_list=FILENAMES)
|
|
m@16
|
80 max_i = np.argmax(df_results[:, 1])
|
|
m@16
|
81 feat_learning_i = max_i % 4 # 4 classifiers for each feature learning method
|
|
m@16
|
82 filename = FILENAMES[feat_learning_i]
|
|
m@16
|
83 X, Y, Yaudio = load_data_from_pickle(filename)
|
|
m@16
|
84 traininds, testinds = get_train_test_indices()
|
|
m@16
|
85 X_train, Y_train, X_test, Y_test = get_train_test_sets(X, Y, traininds, testinds)
|
|
m@16
|
86 confusion_matrix(X_train, Y_train, X_test, Y_test, saveCF=True, plots=True)
|
|
m@16
|
87
|
