m@16: # -*- coding: utf-8 -*-
m@16: """
m@16: Created on Thu Nov 10 15:10:32 2016
m@16: 
m@16: @author: mariapanteli
m@16: """
m@16: import numpy as np
m@16: import pandas as pd
m@16: from sklearn import metrics
m@16: 
m@16: import map_and_average
m@16: import util_feature_learning
m@16: 
m@16: 
m@16: FILENAMES = map_and_average.OUTPUT_FILES
m@16: 
m@16: 
m@16: def load_data_from_pickle(filename):
m@16:     X_list, Y, Yaudio = pickle.load(open(filename,'rb'))
m@16:     X = np.concatenate(data_list, axis=1)
m@16:     return X, Y, Yaudio
m@16: 
m@16: 
m@16: def get_train_test_indices():
m@16:     trainset, valset, testset = map_and_average.load_train_val_test_sets()
m@16:     trainaudiolabels, testaudiolabels = trainset[2], testset[2]
m@16:     # train, test indices
m@16:     aa_train = np.unique(trainaudiolabels)
m@16:     aa_test = np.unique(testaudiolabels)
m@16:     traininds = np.array([i for i, item in enumerate(audiolabs) if item in aa_train])
m@16:     testinds = np.array([i for i, item in enumerate(audiolabs) if item in aa_test])
m@16:     return traininds, testinds
m@16: 
m@16: 
m@16: def get_train_test_sets(X, Y, traininds, testinds):
m@16:     X_train = X[traininds, :]
m@16:     Y_train = Y[traininds]
m@16:     X_test = X[testinds, :]
m@16:     Y_test = Y[testinds]
m@16:     return X_train, Y_train, X_test, Y_test
m@16: 
m@16: 
m@16: def classify_for_filenames(file_list=FILENAMES):
m@16:     df_results = pd.DataFrame()
m@16:     feat_learner = util_feature_learning.Transformer()
m@16:     for filename in file_list:
m@16:         X, Y, Yaudio = load_data_from_pickle(filename)
m@16:         traininds, testinds = get_train_test_indices()
m@16:         X_train, Y_train, X_test, Y_test = get_train_test_sets(X, Y, traininds, testinds)
m@16:         df_result = feat_learner.classify(X_train, Y_train, X_test, Y_test)
m@16:         df_results = pd.concat([df_results, df_result], axis=0, ignore_index=True)
m@16:     return df_results    
m@16: 
m@16: 
m@16: def plot_CF(CF, labels=None, figurename=None):
m@16:     labels[labels=='United States of America'] = 'United States Amer.'
m@16:     plt.imshow(CF, cmap="Greys")
m@16:     plt.xticks(range(len(labels)), labels, rotation='vertical', fontsize=4)
m@16:     plt.yticks(range(len(labels)), labels, fontsize=4)
m@16:     if figurename is not None:
m@16:         plt.savefig(figurename, bbox_inches='tight')
m@16: 
m@16: 
m@16: def confusion_matrix(X_train, Y_train, X_test, Y_test, saveCF=False, plots=False):
m@16:     feat_learner = util_feature_learning.Transformer()
m@16:     accuracy, predictions = util_feature_learning.classification_accuracy(X_train, Y_train, 
m@16:                         X_test, Y_test, model=util_feature_learning.modelLDA)
m@16:     labels = np.unique(Y_test)  # TODO: countries in geographical proximity
m@16:     CF = metrics.confusion_matrix(Y_test, predictions, labels=labels)
m@16:     if saveCF:
m@16:         np.savetxt('data/CFlabels.csv', labels, fmt='%s')
m@16:         np.savetxt('data/CF.csv', CF, fmt='%10.5f')
m@16:     if plots:
m@16:         plot_CF(CF, labels=labels, figurename='data/conf_matrix.pdf')
m@16:     return accuracy, predictions
m@16: 
m@16: 
m@16: if __name__ == '__main__':
m@16:     df_results = classify_for_filenames(file_list=FILENAMES)
m@16:     max_i = np.argmax(df_results[:, 1])
m@16:     feat_learning_i = max_i % 4  # 4 classifiers for each feature learning method
m@16:     filename = FILENAMES[feat_learning_i]
m@16:     X, Y, Yaudio = load_data_from_pickle(filename)
m@16:     traininds, testinds = get_train_test_indices()
m@16:     X_train, Y_train, X_test, Y_test = get_train_test_sets(X, Y, traininds, testinds)
m@16:     confusion_matrix(X_train, Y_train, X_test, Y_test, saveCF=True, plots=True)
m@16: